Forgotten readme edit

Picked from 71e5c5b and 38029e3

.gitmodules

@@ -1,3 +1,6 @@
 [submodule "ucwmac"]
 	path = ucwmac
 	url = git://git.ucw.cz/ucwmac.git
+[submodule "src/formatitko/katex-server"]
+	path = src/formatitko/katex-server
+	url = https://gitea.ks.matfyz.cz:/KSP/formatitko-katex-server

README.md

@@ -16,7 +16,9 @@ Inkscape are used for image processing. Nodejs is used for KaTeX.
 
 ## Usage
 ```
-usage: formatitko.py [-h] [-l IMG_LOOKUP_DIRS [IMG_LOOKUP_DIRS ...]] [-p IMG_PUBLIC_DIR] [-i IMG_WEB_PATH] [-w OUTPUT_HTML] [-t OUTPUT_TEX] input_filename
+usage: formatitko [-h] [-l IMG_LOOKUP_DIRS [IMG_LOOKUP_DIRS ...]] [-p IMG_PUBLIC_DIR] [-c IMG_CACHE_DIR] [-i IMG_WEB_PATH] [-w OUTPUT_HTML] [-t OUTPUT_TEX] [-m OUTPUT_MD]
+                  [-j OUTPUT_JSON] [--katex-server] [-k KATEX_SOCKET] [--debug]
+                  input_filename
 
 positional arguments:
   input_filename        The markdown file to process.
@@ -24,16 +26,27 @@ positional arguments:
 options:
   -h, --help            show this help message and exit
   -l IMG_LOOKUP_DIRS [IMG_LOOKUP_DIRS ...], --img-lookup-dirs IMG_LOOKUP_DIRS [IMG_LOOKUP_DIRS ...]
-                        Image lookup directories. When processing images, the program will try to find the image in them first. Always looks for images in the same folder as the markdown
-                        file. (default: [])
+                        Image lookup directories. When processing images, the program will try to find the image in them first. Always looks for images in the same folder
+                        as the markdown file. (default: [])
   -p IMG_PUBLIC_DIR, --img-public-dir IMG_PUBLIC_DIR
-                        Directory to put processed images into. The program will not overwrite existing images. (default: public)
+                        Directory to put processed images into. The program will overwrite images, whose dependencies are newer. (default: public)
+  -c IMG_CACHE_DIR, --img-cache-dir IMG_CACHE_DIR
+                        Directory to cache processed images and intermediate products. The program will overwrite files, whose dependencies are newer. (default: cache)
   -i IMG_WEB_PATH, --img-web-path IMG_WEB_PATH
                         Path where the processed images are available on the website. (default: /)
   -w OUTPUT_HTML, --output-html OUTPUT_HTML
-                        The HTML file (for Web) to write into. (default: output.html)
+                        The HTML file (for Web) to write into. (default: None)
   -t OUTPUT_TEX, --output-tex OUTPUT_TEX
-                        The TEX file to write into. (default: output.tex)
+                        The TEX file to write into. (default: None)
+  -m OUTPUT_MD, --output-md OUTPUT_MD
+                        The Markdown file to write into. (Uses pandoc to generate markdown) (default: None)
+  -j OUTPUT_JSON, --output-json OUTPUT_JSON
+                        The JSON file to dump the pandoc-compatible AST into. (default: None)
+  --katex-server        Starts a KaTeX server and prints the socket filename onto stdout. Useful for running formatitko many times without starting the KaTeX server each
+                        time. (default: False)
+  -k KATEX_SOCKET, --katex-socket KATEX_SOCKET
+                        The KaTeX server socket filename obtained by running with `--katex-server`. (default: None)
+  --debug
 ```
 
 ## Format
@@ -46,7 +59,7 @@ definition lists and citations. It also adds its own custom features.
 
 Flags can be set in the Front Matter or with python code. Then, elements with
 the `if` attribute will only be shown if the flag is set to True and elements
-with the `ifn` attribute will only be show if the flag is not set to True.
+with the `ifnot` attribute will only be show if the flag is not set to True.
 
 **Example:**
 
@@ -59,7 +72,7 @@ flags:
 
 [This will not be shown]{if=bar}
 
-[This will be shown]{ifn=bar}
+[This will be shown]{ifnot=bar}
 ```
 
 ### Including other files
@@ -69,12 +82,28 @@ There are two ways of including files.
 #### Importing
 The first is importing, which only takes the state (defined commands, metadata,
 etc.) from the file and any content is omitted. This is useful for creating
-libraries of commands. The syntax is as follows:
+libraries of commands.
 
-[#test/empty.md]{}
+There are three types of imports:
 
-The curly braces are required for pandoc to parse the import properly and should
-be left empty.
+##### Python Module (the default)
+```markdown
+[#ksp_formatitko as ksp]{}
+```
+or
+```markdown
+[#ksp_formatitko]{}
+```
+with an optional `type=module` in the curly brackets, tries to import a python
+module as a set of formatitko commands. See below for more details about
+commands.
+
+##### JSON Metadata
+[#test/test.json]{type=metadata key=orgs}
+
+This will import metadata from a JSON file. THe optional `key` argument sets the
+key under which the whole JSON file will be placed. Dictionaries are merged,
+others overwritten.
 
 #### Partials
 Partials are the very opposite of imports, they have their own context, which
@@ -95,12 +124,19 @@ partial to `tex` or `html`.
 ### Groups
 
 Groups are pieces of markdown with their own sandboxed context, in other words,
-inline partials. They function exactly the same as partials, namely can have
-their own front matter.
+inline partials. Syntax-wise they are pandoc Divs with the `.group` class. All
+attributes of the Div will be passed down as metadata to the group.
+
+::: {.group lang=cs}
+OOOoo český mód
+:::
+
+If you want to have more fancy metadata, that can only be specified in a front
+matter, you can use the following syntax:
 
 ```markdown {.group}
 ---
-language: cs
+lang: cs
 ---
 OOOoo český mód
 ```
@@ -114,6 +150,9 @@ fmt.Pritln("owo")
 ```
 ````
 
+Note however, that when this syntax is used, pandoc is executed for each of
+these blocks which could get slow. Using divs is preferred.
+
 Groups and partials are also enclosed in `\begingroup` and `\endgroup` in the
 output TeX.
 
@@ -138,15 +177,38 @@ pandoc feature.]
 ### Running python code
 
 Formátítko allows you to run Python code directly from your MD file. Any
-`python` code block with the class `run` will be executed:
+`python` code block with the class `run` will be executed.
 
-#### Context
+#### Command environment
 
-You can access the current context using the `ctx` variable. The context
+The commands will be executed as functions with the following signature:
+```python
+def command(element: Command, context: Context) -> list[Element]:
+```
+some global variables may be available, and are defined in `command_env.py`:
+```python
+import panflute as pf
+import formatitko.elements as fe
+from formatitko.util import import_md_list
+from formatitko.util import parse_string
+
+from formatitko.context import Context
+from formatitko.command import Command
+from panflute import Element
+```
+##### `element` parameter
+
+The `element` parameter holds the element the command is currently being executed
+on. In the case of running python blocks directly, it is probably not
+interesting but will get interesting later.
+
+##### `context` parameter
+
+You can access the current context using the `context` parameter. The context
 provides read/write access to the FrontMatter metadata. The context has the
 following methods:
 
-`ctx.get_metadata(key: str, simple: bool=True, immediate: bool=False)`
+`context.get_metadata(key: str, simple: bool=True, immediate: bool=False)`
 
 - `key`: The key of the metadatum you want to get. Separate child keys with
   dots: `ctx.get_metadata("flags.foo")`
@@ -156,13 +218,13 @@ following methods:
 - `immediate`: Only get metadatum from the current context, not from its
   parents.
 
-`ctx.set_metadata(key: str, value)`
+`context.set_metadata(key: str, value)`
 
 - `key`: The key of the metadatum you want to get. Separate child keys with
   dots: `ctx.get_metadata("flags.foo")`
 - `value`: Any value you want to assign to the metadatum
 
-`ctx.unset_metadata(key: str)`
+`context.unset_metadata(key: str)`
 
 Delete the metadatum in the current context and allow it to inherit the value
 from the parent context.
@@ -172,26 +234,31 @@ from the parent context.
 
 Helper functions for flags exist which work the same as for metadata:
 
-`ctx.is_flag_set(flag: str) -> bool`
+`context.is_flag_set(flag: str) -> bool`
 
-`ctx.set_flag(flag: str, val: bool)`
+`context.set_flag(flag: str, val: bool)`
 
-`ctx.unset_flag(flag: str)`
+`context.unset_flag(flag: str)`
 
-#### Writing output
+There are also other useful functions, which you can see for yourself in
+`context.py`.
 
-There are two modes of writing output, plaintext and element-based.
+> **WARNING**: Writing to metadata should **only** be done **at the beginning**
+> of the document or a group (before any printable content). Writing to metadata
+> in other places in the document might cause undefined behaviour (mostly some
+> elements might behave as if the metadata was set elsewhere).
 
-Plaintext mode uses the `print(text: str)` and `println(text: str)` functions,
-that append text to a buffer which is then interpreted as markdown input.
+##### Return value
+The function **must** return a list of valid Elements. This list may be empty.
+These elements will be placed in the document in the location where the command
+was invoked.
 
-Element-based mode uses the `appendChild(element: pf.Element)` and
-`appendChildren(*elements: List[pf.Element])` functions which allow you to
-append `panflute` elements to a list which is then again interpreted as input.
-The `panflute` library is available as `pf`.
+The `parse_string` function might be useful, it turns a simple string into a
+list of panflute's `Str`s and `Space`s (without any formatting). If you want to
+use markdown in your function output, you have to convert it yourself using
+`import_md` but beware this calls pandoc, is potentially slow and is
+discouraged.
 
-When one of these functions is called, the mode is set and functions from the
-other mode cannot be called within the same block of code.
 
 **Examples:**
 
@@ -200,14 +267,15 @@ other mode cannot be called within the same block of code.
 title: Foo
 ---
 ```python {.run}
-println("*wooo*")
-println()
-println("The title of this file is: " + ctx.get_metadata("title"))
+return [
+	pf.Para(pf.Emph(pf.Str("wooo"))),
+	pf.Para(*parse_string("The title of this file is: " + context.get_metadata("title")))
+]
 ```
 ````
 
 ```python {.run}
-appendChild(pf.Para(pf.Strong(pf.Str("foo"))))
+return [pf.Strong(*parse_string("Hello world!"))]
 ```
 
 ### Defining and running commands
@@ -218,7 +286,7 @@ Code blocks can be also saved and executed later. Defining is done using the
 **Example:**
 
 ```python {define=commandname}
-print("foo")
+return [pf.Str("foo")]
 ```
 
 If you try to define the same command twice, you will get an error. To redefine
@@ -230,7 +298,7 @@ There are multiple ways of running commands. There is the shorthand way:
 
 [!commandname]{}
 
-Or using the `c` attribute on a span or a div:
+Or using the `c` attribute on a span or a div (new: or a codeblock!):
 
 [Some content]{c=commandname}
 
@@ -238,6 +306,16 @@ Or using the `c` attribute on a span or a div:
 Some content
 :::
 
+```python {define=bash}
+import subprocess
+c = subprocess.run(["bash", "-c", element.text], stdout=subprocess.PIPE, check=True, encoding="utf-8")
+return [pf.Para(pf.Str(c.stdout))]
+```
+
+```bash {c=bash}
+cat /etc/hostname
+```
+
 To access the content or attributes of the div or span the command has been
 called on, the `element` variable is available, which contains the `panflute`
 representation of the element.
@@ -245,7 +323,7 @@ representation of the element.
 **Example:**
 
 ```python {define=index}
-appendChild(element.content[int(element.attributes["i"])])
+return [element.content[int(element.attributes["i"])]]
 ```
 
 [Pick the third element from this span]{c=index i=2}
@@ -268,23 +346,24 @@ blocks. To turn it off for a single block, don't specify a language or set the
 `highlight` attribute to `False`. You can also set the metadatum `highlight` to
 `false` in the FrontMatter to disable it in a given Group. To change the [highlighting
 style](https://pygments.org/styles/), you have to set the `highlight-style`
-metadatum in the **top-level document** this is to prevent the need for many
-inline style definitions.
+metadatum or the `style` attribute directly on the element.
 
 **Examples:**
 ```python
 print("cool")
 ```
 
+```python {style=manni}
+print("freezing")
+```
+
 ```zsh {highlight=False}
 ./formatitko.py README.md
 ```
 
 ### Language awareness
-Formátítko is language aware, this means that the `language` metadatum is
-somewhat special. When set using the front matter, it is also popped out to TeX
-as a `\languagexx` macro. Currently supported values are `cs` and `en` for
-internal uses but can be set to anything.
+Formátítko is language aware, this means that the `lang` metadatum is
+somewhat special. (It is also special for pandoc)
 
 ### NBSP
 Formátítko automatically inserts no-break spaces according to its sorta smart
@@ -303,12 +382,9 @@ language.
 
 **Examples:**
 
-```markdown {.group}
----
-language: cs
----
+::: {.group lang=cs}
 "Uvozovky se v českém testu píší 'jinak' než v angličtině."
-```
+:::
 
 "In Czech texts, quotes are written 'differently' than in English"
 
@@ -339,6 +415,9 @@ Images are automatically searched for in the directory where each markdown file
 command line parameter. After processing, they're all put into the folder
 specified with `--public-dir`.
 
+Formátítko also does dependency management, which means that all images will be
+regenerated only when their dependencies are newer.
+
 #### Image processing
 Images are automatically processed so that they can be successfully used in both
 output formats. This includes generating multiple sizes and providing a
@@ -348,12 +427,22 @@ To customize this, the `file-width`, `file-height`, `file-dpi`, `file-quality`
 and `no-srcset` attributes are available. All but the last one should be
 integers.
 
-Keep in mind that the processing tries to be as lazy as possible, so it never
-overwrites any files and if it finds the right format or resolution (only
-judging by the filenames) in the lookup directories it will just copy that. This
-means that any automatic attempts at conversion can be overridden by converting
-the file yourself, naming it accordingly and placing it either in the public or
-one of the lookup directories.
+Processing also includes Asymptote images -- you can simply include an asymptote
+program as an image and formátítko handles the rest for you.
+
+#### Content headers and footers
+
+If you want formatitko to generate fully formed html files for you, you might
+want to add a HTML partial with the starting tags and `<head>`. This would
+normally not work, because the entire document is wrapped with `<main>`. Using
+the special `.header_content` and `.footer_content` classes of divs, you can
+append content to a header and footer, which are popped to the output before and
+after the document.
+
+:::: {.header_content}
+::: {partial="test/test-top.html" type="html"}
+:::
+::::
 
 ## Working with the produced output
 
@@ -366,11 +455,15 @@ your `<head>`^[This is taken directly from [KaTeX's docs](https://katex.org/docs
 <link rel='stylesheet' href='https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css' integrity='sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0' crossorigin='anonymous'>
 ```
 
-Also the output HTML is not intended as a standalone file but should be included
-as part of a larger template. (That includes a doctype, other css, etc.)
+You can see how this is done in `test/test.md`
 
 ### TeX
 The TeX output is not usable as is. Many of the elements are just converted to
 macros, which you have to define yourself. There is an example implementation in
 `formatitko.tex`, which uses LuaTeX and the ucwmac package, but you should
 customize it to your needs (and to the context in which the output is used).
+
+## More examples
+
+More usage examples can be found (even though a bit chaotically) in the test
+directory.

formatitko.tex

@@ -1,9 +1,6 @@
 \input luatex85.sty
 \input ucwmac2.tex
-\ucwmodule{luaofs}
-\ucwmodule{link}
-\ucwmodule{verb}
-\parskip=3pt plus 2pt minus 1pt
+\parskip=5pt plus 3pt minus 2pt
 \parindent=0sp
 
 \def\strong#1{{%
@@ -17,23 +14,17 @@
 
 
 \def\superscript#1{\leavevmode\raise3pt\hbox{\fiverm#1}}
+\def\subscript#1{\leavevmode\lower1pt\hbox{\fiverm#1}}
 \newcount\fncount
 \fncount=1
 \def\fnmark{\superscript{\the\fncount}}
 \def\fn#1{\footnote\fnmark{#1}\advance\fncount by 1}
-\def\hA#1{{\parskip1em\settextsize{14}\bf #1}}
-\def\hB#1{{\parskip1em\settextsize{12}\bf #1}}
-\def\hC#1{{\parskip1em\settextsize{10}\bf #1}}
-\def\hD#1{{\parskip1em\settextsize{10}\bi #1}}
-\def\hr{{\vskip5pt\hrule\vskip5pt}}
+\def\section#1{{\parskip1em\settextsize{18}\bf #1}}
+\def\subsection#1{{\parskip1em\settextsize{16}\bf #1}}
+\def\subsubsection#1{{\parskip1em\settextsize{14}\bf #1}}
+\def\subsubsubsection#1{{\parskip1em\settextsize{12}\bf #1}}
+\def\subsubsubsubsection#1{{\parskip1em\settextsize{10}\bf #1}}
+\def\subsubsubsubsubsection#1{{\parskip1em\settextsize{10}\bi #1}}
 \long\def\blockquote#1{\vskip\lineskip\vskip\parskip\hbox{\vrule\hskip5pt\vbox{#1}}}
-\let\code\verbatim
-\let\codeblock\verbatim
-\def\subscript#1{\leavevmode\lower1pt\hbox{\fiverm#1}}
 \def\strikeout#1{FIXME: Strikeout not implemented}
 \def\underline#1{FIXME: Underline not implemented}
-\def\figure#1#2{\vskip5pt\centerline{#1}\centerline{#2}\vskip5pt}
-\def\figcaption#1{{\it #1}}
-\let\image\putimage
-\def\languagecs{} % KSP should define this to \cze probably
-\def\languageen{} % KSP should define this to \eng probably

src/formatitko/command.py

@@ -16,7 +16,7 @@ class InlineCommand(Span, Command):
 			if len(content) == 1 and (isinstance(content[0], Para) or isinstance(content[0], Plain)):
 				return Span(*content[0].content)
 			else:
-				raise InlineError(f"The command {self.attributes['c']} returned multiple Paragraphs and must be executed using `::: {{c={self.attributes['c']}}}\\n:::`.\n\n{content}")
+				return Div(*content)
 	pass
 
 class BlockCommand(Div, Command):

src/formatitko/command_env.py

@@ -5,4 +5,5 @@ from formatitko.util import parse_string
 
 from formatitko.context import Context
 from formatitko.command import Command
+from .nop_processor import NOPProcessor
 from panflute import Element

src/formatitko/command_util.py

@@ -15,10 +15,10 @@ def parse_command(code: str) -> CommandCallable:
 	indented_code_lines = []
 	for line in code_lines:
 		indented_code_lines.append(("\t" if tabs else "    ")+line)
-	code = "def command(element: Command, context: Context) -> list[Element]:\n"+"\n".join(indented_code_lines)
-	globals = command_env.__dict__
-	exec(code, globals)
-	return globals["command"]
+	code = "def command(element: Command, context: Context, processor: NOPProcessor) -> list[Element]:\n"+"\n".join(indented_code_lines)
+	env = {**command_env.__dict__}
+	exec(code, env)
+	return env["command"]
 
 # This function is called in trasform.py, defining a command which can be
 # called later

src/formatitko/context.py

@@ -3,11 +3,10 @@ from panflute import Doc, Element, Div, Span
 from typing import Union, Callable
 from types import ModuleType
 import os
-import warnings
 
 from .command import Command
 
-CommandCallable = Callable[[Command, 'Context'], list[Element]] # This is here because of a wild circular import dependency between many functions and classes
+CommandCallable = Callable[[Command, 'Context', 'NOPProcessor'], list[Element]] # This is here because of a wild circular import dependency between many functions and classes
 
 # This class is used to keep state while transforming the document using
 # transform.py. For the context to be available to the html and TeX generators,
@@ -23,20 +22,29 @@ CommandCallable = Callable[[Command, 'Context'], list[Element]] # This is here b
 class Context:
 	parent: Union["Context", None]
 	_commands: dict[str, Union[CommandCallable, None]]
+	_data: dict[str, object]
 	doc: Doc
 	trusted: bool
 	path: str
 	dir: str
 	filename: str
+	root_dir: str  # Absolute path to the dir of the file formátítko was called on
+	rel_dir: str   # Relative path to the current dir from the root dir
+	deps: set[str]
 
 	def __init__(self, doc: Doc, path: str, parent: Union['Context', None]=None, trusted: bool=True):
 		self.parent = parent
 		self._commands = {}
+		self._data = {}
 		self.doc = doc
 		self.trusted = trusted
 		self.path = path
 		self.dir = os.path.dirname(path) if os.path.dirname(path) != "" else "."
 		self.filename = os.path.basename(path)
+		self.root_dir = parent.root_dir if parent else os.path.abspath(self.dir)
+		self.rel_dir = os.path.relpath(self.dir, self.root_dir)
+		self.deps = set()
+		self.add_dep(path)
 		if self.get_metadata("flags", immediate=True) is None:
 			self.set_metadata("flags", {})
 
@@ -63,7 +71,7 @@ class Context:
 				self.set_command(prefix+name, func)
 
 	def is_flag_set(self, flag: str):
-		if self.get_metadata("flags."+flag):
+		if self.get_metadata("flags."+flag) is not None:
 			if self.get_metadata("flags."+flag):
 				return True
 			else:
@@ -109,7 +117,57 @@ class Context:
 		else:
 			self.set_metadata(key, data)
 
+	def get_data(self, key: str, immediate: bool=False):
+		data = self._data
+		keys = key.split(".")
+		try:
+			for k in keys:
+				data = data[k]
+			return data
+		except KeyError:
+			if self.parent and not immediate:
+				return self.parent.get_data(key)
+			else:
+				return None
 
+	def set_data(self, key: str, value: object):
+		data = self._data
+		keys = key.split(".")
+		for k in keys[:-1]:
+			try:
+				data = data[k]
+			except KeyError:
+				data[k] = {}
+				data = data[k]
+		data[keys[-1]] = value
+
+	def unset_data(self, key: str):
+		if key == "":
+			self._data = {}
+		data = self._data
+		keys = key.split(".")
+		for k in keys[:-1]:
+			data = data[k]
+		del data[keys[-1]]
+
+	def get_deps(self) -> list[str]:
+		if self.parent is not None:
+			return self.parent.get_deps()
+		else:
+			return self.deps
+
+	def add_dep(self, dep: str):
+		self.get_deps().add(os.path.abspath(dep))
+
+	def add_deps(self, deps: list[str]):
+		self.get_deps().update([os.path.abspath(path) for path in deps])
+
+
+def get_context_from_doc(doc: Doc) -> Context:
+	if len(doc.content) == 1 and isinstance(doc.content[0], Group):
+		return doc.content[0].context
+	else:
+		return None
 
 # This is a custom element which creates \begingroup \endgroup groups in TeX
 # and also causes KaTeX math blocks to be isolated in a similar way.

src/formatitko/elements.py

@@ -1,4 +1,4 @@
-from panflute import Quoted
+from panflute import Quoted, Link
 
 
 from .command import Command, InlineCommand, BlockCommand, CodeCommand
@@ -14,3 +14,6 @@ class FQuoted(Quoted):
 		del kwargs["style"]
 		super().__init__(*args, **kwargs)
 
+
+class FileLink(Link):
+	pass

src/formatitko/formatitko.py

@@ -2,19 +2,20 @@
 
 import argparse
 import sys
+import tempfile
+import subprocess
+import shutil
 
 # Import local files
-from .transform import transform
 from .util import import_md
-from .context import Context, BlockGroup
 from .katex import KatexClient
-from .html import html
-from .tex import tex
-from .images import ImageProcessor
-from .output_generator import OutputGenerator
-from .html_generator import HTMLGenerator
+from .images import ImageProcessor, ImageProcessorNamespace
+from .output_generator import OutputGenerator, FormatitkoRecursiveError
+from .html_generator import HTMLGenerator, StandaloneHTMLGenerator
 from .transform_processor import TransformProcessor
 from .pandoc_processor import PandocProcessor
+from .tex_generator import UCWTexGenerator
+from .context import get_context_from_doc
 
 from panflute import convert_text
 
@@ -26,13 +27,17 @@ def main():
 	parser.add_argument("-c", "--img-cache-dir", help="Directory to cache processed images and intermediate products. The program will overwrite files, whose dependencies are newer.", default="cache")
 	parser.add_argument("-i", "--img-web-path", help="Path where the processed images are available on the website.", default="/")
 	parser.add_argument("-w", "--output-html", help="The HTML file (for Web) to write into.")
+	parser.add_argument("-s", "--output-standalone-html", help="The Standalone HTML file to write into. A full page is generated instead of just a fragment.")
 	parser.add_argument("-t", "--output-tex", help="The TEX file to write into.")
 	parser.add_argument("-m", "--output-md", help="The Markdown file to write into. (Uses pandoc to generate markdown)")
 	parser.add_argument("-j", "--output-json", help="The JSON file to dump the pandoc-compatible AST into.")
+	parser.add_argument("-P", "--output-pdf", help="The PDF file to save the PDF generated by TeX.")
 	parser.add_argument("--katex-server", action='store_true', help="Starts a KaTeX server and prints the socket filename onto stdout. Useful for running formatitko many times without starting the KaTeX server each time.")
 	parser.add_argument("-k", "--katex-socket", help="The KaTeX server socket filename obtained by running with `--katex-server`.")
 	parser.add_argument("input_filename", help="The markdown file to process.", nargs="?" if "--katex-server" in sys.argv else None)
 	parser.add_argument("--debug", action='store_true')
+	parser.add_argument("--traceback-limit", help="Traceback limit for when errors happen, defaults to 0, as it is only useful for internal debugging.", default=0)
+	parser.add_argument("--deps", help="File to write list of dependencies to. May depend on output formats used.")
 	args = parser.parse_args()
 
 	if args.katex_server:
@@ -49,27 +54,78 @@ def main():
 	doc = import_md(open(args.input_filename, "r").read())
 
 	if args.debug:
-		OutputGenerator(sys.stdout).generate(doc)
+		try:
+			OutputGenerator(sys.stdout).generate(doc)
+		except FormatitkoRecursiveError as e:
+			e.pretty_print(tracebacklimit=args.traceback_limit)
 
-	doc = TransformProcessor(args.input_filename).transform(doc)
+	try:
+		doc = TransformProcessor(args.input_filename).transform(doc)
+	except FormatitkoRecursiveError as e:
+		e.pretty_print(tracebacklimit=args.traceback_limit)
 
 	# Initialize the image processor (this just keeps some basic state)
-	imageProcessor = ImageProcessor(args.img_public_dir, args.img_web_path, args.img_cache_dir, *args.img_lookup_dirs)
+	imageProcessor = ImageProcessor({"": ImageProcessorNamespace(args.img_public_dir, args.img_web_path, args.img_cache_dir, args.img_lookup_dirs, True)})
 
 	if args.output_html is not None:
 		# Initialize KaTeX client (this runs the node app and connects to a unix socket)
 		with KatexClient(socket=args.katex_socket) as katexClient:
-			HTMLGenerator(open(args.output_html, "w"), katexClient, imageProcessor).generate(doc)
+			with open(args.output_html, "w") as file:
+				try:
+					HTMLGenerator(file, katexClient, imageProcessor).generate(doc)
+				except FormatitkoRecursiveError as e:
+					e.pretty_print(tracebacklimit=args.traceback_limit)
+
+	if args.output_standalone_html is not None:
+		# Initialize KaTeX client (this runs the node app and connects to a unix socket)
+		with KatexClient(socket=args.katex_socket) as katexClient:
+			with open(args.output_standalone_html, "w") as file:
+				try:
+					StandaloneHTMLGenerator(file, katexClient, imageProcessor).generate(doc)
+				except FormatitkoRecursiveError as e:
+					e.pretty_print(tracebacklimit=args.traceback_limit)
+
+	if args.output_tex is not None:
+		with open(args.output_tex, "w") as file:
+			try:
+				UCWTexGenerator(file, imageProcessor).generate(doc)
+			except FormatitkoRecursiveError as e:
+				e.pretty_print(tracebacklimit=args.traceback_limit)
 
 	if args.output_md is not None:
-		open(args.output_md, "w").write(convert_text(PandocProcessor().transform(doc), input_format="panflute", output_format="markdown"))
+		with open(args.output_md, "w") as file:
+			file.write(convert_text(PandocProcessor().transform(doc), input_format="panflute", output_format="markdown"))
 
 	if args.output_json is not None:
-		open(args.output_json, "w").write(convert_text(PandocProcessor().transform(doc), input_format="panflute", output_format="json"))
+		with open(args.output_json, "w") as file:
+			file.write(convert_text(PandocProcessor().transform(doc), input_format="panflute", output_format="json"))
+
+	if args.output_pdf is not None:
+		if args.output_tex is None:
+			fd = tempfile.NamedTemporaryFile(dir=".", suffix=".tex")
+			with open(fd.name, "w") as file:
+				try:
+					UCWTexGenerator(file, imageProcessor).generate(doc)
+				except FormatitkoRecursiveError as e:
+					e.pretty_print(tracebacklimit=args.traceback_limit)
+			filename = fd.name
+		else:
+			filename = args.output_tex
+		outdir = tempfile.TemporaryDirectory(prefix="formatitko")
+		subprocess.run(["pdfcsplain", "-halt-on-error", "-output-directory="+outdir.name, "-jobname=formatitko", filename], check=True)
+		shutil.move(outdir.name+"/formatitko.pdf", args.output_pdf)
+
+	if args.deps is not None:
+		with open(args.deps, "w") as file:
+			for dep in get_context_from_doc(doc).get_deps():
+				file.write(dep + "\n")
 
 	if args.debug:
 		print("-----------------------------------")
-		OutputGenerator(sys.stdout).generate(doc)
+		try:
+			OutputGenerator(sys.stdout).generate(doc)
+		except FormatitkoRecursiveError as e:
+			e.pretty_print(tracebacklimit=args.traceback_limit)
 
 
 if __name__ == "__main__":

src/formatitko/html.py

@@ -1,311 +0,0 @@
-from panflute import *
-from pygments import highlight
-from pygments.lexers import get_lexer_by_name
-from pygments.formatters import HtmlFormatter
-from pygments.util import ClassNotFound
-import os
-from typing import Union
-
-from .whitespace import NBSP
-from .elements import FQuoted
-from .katex import KatexClient
-from .util import inlinify
-from .context import Group
-from .images import ImageProcessor
-
-import warnings
-warnings.warn("The html function has been deprecated, is left only for reference and will be removed in future commits. HTML_generator should be used in its place.", DeprecationWarning)
-
-def html(e: Union[Element, ListContainer], k: KatexClient, i: ImageProcessor, indent_level: int=0, indent_str: str="\t") -> str:
-
-	warnings.warn("The html function has been deprecated, is left only for reference and will be removed in future commits. HTML_generator should be used in its place.", DeprecationWarning)
-
-	# `only` attribute which makes transformed elements appear only in tex
-	# output or html output
-	if hasattr(e, "attributes") and "only" in e.attributes and e.attributes["only"] != "html":
-		return ""
-
-	if isinstance(e, ListContainer):
-		return ''.join([html(child, k, i, indent_level, indent_str) for child in e])
-
-	# Bits from which the final element output is built at the end of this
-	# function. Most elements override this by returning their own output.
-	tag = e.tag.lower()
-	attributes = ""
-	content_foot = ""
-	content_head = ""
-
-	if isinstance(e, Str):
-		return e.text.replace(" ", " ")
-
-	# Most elements fit the general template at the end of the function, just
-	# need their html tag specified.
-	tags = {
-		BulletList: "ul",
-		Doc: "main",
-		Emph: "em",
-		Caption: "figcaption",
-		Para: "p",
-		Header: "h"+str(e.level) if isinstance(e, Header) else "",
-		LineBlock: "p",
-		ListItem: "li",
-		SmallCaps: "span",
-		Strikeout: "strike",
-		Subscript: "sub",
-		Superscript: "sup",
-		Underline: "u",
-		TableBody: "tbody",
-		TableHead: "thead",
-		TableFoot: "tfoot",
-		TableRow: "tr",
-		TableCell: "td",
-	}
-	if type(e) in tags:
-		tag = tags[type(e)]
-
-	# These are also disabled in pandoc so they shouldn't appear in the AST at all.
-	not_implemented = {
-		Citation: True,
-		Cite: True,
-		Definition: True,
-		DefinitionItem: True,
-		DefinitionList: True
-	}
-	if type(e) in not_implemented:
-		return f'<!-- FIXME: {type(e)}s not implemented -->'
-
-	# Elements which can be represented by a simple string
-	simple_string = {
-		NBSP: "&nbsp;",
-		Space: " ",
-		Null: "",
-		LineBreak: f"\n{indent_level*indent_str}<br>\n{indent_level*indent_str}",
-		SoftBreak: f" ",
-		HorizontalRule: f"{indent_level*indent_str}<hr>\n"
-	}
-	if type(e) in simple_string:
-		return simple_string[type(e)]
-
-	if hasattr(e, "identifier") and e.identifier != "":
-		attributes += f' id="{e.identifier}"'
-
-	if hasattr(e, "classes") and len(e.classes) != 0:
-		attributes += f' class="{" ".join(e.classes)}"'
-
-	# Attributes are only passed down manually, because we use them internally.
-	# Maybe this should be a blocklist instead of an allowlist?
-
-	# Overriding elements with their own returns
-	if isinstance(e, CodeBlock):
-		if len(e.classes) > 0 and (e.attributes["highlight"] == True or e.attributes["highlight"] == 'True'):
-			# Syntax highlighting using pygments
-			for cl in e.classes:
-				try:
-					lexer = get_lexer_by_name(cl)
-				except ClassNotFound:
-					continue
-				break
-			else:
-				print(f"WARN: Syntax highligher does not have lexer for element with these classes: {e.classes}")
-			formatter = HtmlFormatter(style=e.attributes["style"])
-			result = highlight(e.text, lexer, formatter)
-			return f'{result}'
-		else:
-			return f'<pre>{e.text}</pre>'
-
-	if isinstance(e, Doc):
-		formatter = HtmlFormatter(style=e.get_metadata("highlight-style") if e.get_metadata("highlight-style") is not None else "default")
-		content_head = f'<style>{formatter.get_style_defs(".highlight")}</style>'
-
-	if isinstance(e, Image):
-		url = e.url
-
-		# Attributes → image processor args
-		additional_args = {}
-		if "file-width" in e.attributes:
-			additional_args["width"] = int(e.attributes["file-width"])
-		if "file-height" in e.attributes:
-			additional_args["height"] = int(e.attributes["file-height"])
-		if "file-quality" in e.attributes:
-			additional_args["quality"] = int(e.attributes["file-quality"])
-		if "file-dpi" in e.attributes:
-			additional_args["dpi"] = int(e.attributes["file-dpi"])
-
-		# The directory of the current file, will also look for images there.
-		source_dir = e.attributes["source_dir"]
-
-		_, ext = os.path.splitext(url)
-		ext = ext[1:]
-
-		# Conversions between various formats.
-		if ext in ["svg", "png", "jpeg", "gif"]:
-			# Even supported elements have to be 'converted' because the
-			# processing contains finding and moving them to the output
-			# directory.
-			url = i.process_image(url, ext, source_dir, **additional_args)
-		elif ext in ["pdf", "epdf"]:
-			if not "dpi" in additional_args:
-				additional_args["dpi"] = 300
-			url = i.process_image(url, "png", source_dir, **additional_args)
-		elif ext in ["jpg"]:
-			url = i.process_image(url, "jpeg", source_dir, **additional_args)
-		else:
-			url = i.process_image(url, "png", source_dir, **additional_args)
-		
-		# Srcset generation - multiple alternative sizes of images browsers can
-		# choose from.
-		_, ext = os.path.splitext(url)
-		ext = ext[1:]
-		srcset = []
-		if ext in ["png", "jpeg"] and (not "no-srcset" in e.attributes or e.attributes["no-srcset"] == False or e.attributes["no-srcset"] == 'False'):
-			# This is inspired by @vojta001's blogPhoto shortcode he made for
-			# patek.cz:
-			# https://gitlab.com/patek-devs/patek.cz/-/blob/master/themes/patek/layouts/shortcodes/blogPhoto.html
-			width, height = i.get_image_size(url, [i.public_dir])
-			sizes = [(640, 360, 85), (1280, 720, 85), (1920, 1080, 90)] # (widht, height, quality)
-			for size in sizes:
-				if width <= size[0] and height <= size[1]:
-					srcset.append((f'{i.web_path}/{url}', f'{width}w'))
-					break
-				quality = size[2] if ext == "jpeg" else None
-				srcset.append((f'{i.web_path}/{i.process_image(url, ext, i.public_dir, width=size[0], height=size[1], quality=quality)}', f'{size[0]}w'))
-
-		url = i.web_path + "/" + url
-		
-		attributes = f'{" style=width:"+e.attributes["width"] if "width" in e.attributes else ""} alt="{e.title or html(e.content, k, i, 0, "")}"'
-		if len(srcset) != 0:
-			return f'<a href="{url}"><img src="{srcset[-1][0]}" srcset="{", ".join([" ".join(src) for src in srcset])}"{attributes}></a>'
-		else:
-			return f'<img src="{url}"{attributes}>'
-
-	# See https://pandoc.org/MANUAL.html#line-blocks
-	if isinstance(e, LineItem):
-		return indent_level*indent_str + html(e.content, k, i) + "<br>\n"
-
-	# Footnotes are placed into parentheses. (And not footnotes (This is how KSP did it before me))
-	if isinstance(e, Note):
-		content_head = "("
-		content_foot = ")"
-		if inlinify(e) is not None:
-			return f' <note>({html(inlinify(e), k, i, 0, "")})</note>'
-
-	if isinstance(e, FQuoted):
-		if e.style == "cs":
-			if e.quote_type == "SingleQuote":
-				return f'‚{html(e.content, k, i, 0, "")}‘'
-			elif e.quote_type == "DoubleQuote":
-				return f'„{html(e.content, k, i, 0, "")}“'
-		elif e.style == "en":
-			if e.quote_type == "SingleQuote":
-				return f'‘{html(e.content, k, i, 0, "")}’'
-			elif e.quote_type == "DoubleQuote":
-				return f'“{html(e.content, k, i, 0, "")}”'
-		else:
-			if e.quote_type == "SingleQuote":
-				return f'\'{html(e.content, k, i, 0, "")}\''
-			elif e.quote_type == "DoubleQuote":
-				return f'"{html(e.content, k, i, 0, "")}"'
-			else:
-				return f'"{html(e.content, k, i, 0, "")}"'
-
-	if isinstance(e, Group):
-		k.begingroup()
-		ret = html(e.content, k, i, indent_level, indent_str)
-		k.endgroup()
-		return ret
-
-	if isinstance(e, Math):
-		formats = {
-			"DisplayMath": True,
-			"InlineMath": False
-		}
-		return indent_level*indent_str + k.render(e.text, {"displayMode": formats[e.format]})
-
-	if isinstance(e, RawInline):
-		if e.format == "html":
-			return e.text
-		else:
-			return ""
-
-	if isinstance(e, RawBlock):
-		if e.format == "html":
-			return f'{e.text}\n'
-		else:
-			return ""
-
-
-	# Non-overriding elements, they get generated using the template at the end
-	# of this function
-	if isinstance(e, Header):
-		tag = "h"+str(e.level)
-
-	if isinstance(e, Figure):
-		content_foot = html(e.caption, k, i, indent_level+1, indent_str)
-
-	if isinstance(e, Caption):
-		tag = "figcaption"
-
-	if isinstance(e, Link):
-		tag = "a"
-		attributes += f' href="{e.url}"'
-		if e.title:
-			attributes += f' title="{e.title}"'
-
-	if isinstance(e, OrderedList):
-		tag = "ol"
-		if e.start and e.start != 1:
-			attributes += f' start="{e.start}"'
-		html_styles = {
-			"Decimal": "1",
-			"LowerRoman": "i",
-			"UpperRoman:": "I",
-			"LowerAlpha": "a",
-			"UpperAlpha": "A"
-		}
-		if e.style and e.style != "DefaultStyle":
-			attributes += f' type="{html_styles[e.style]}"'
-		# FIXME: Delimeter styles
-
-	if isinstance(e, Table):
-		content_head = html(e.head, k, i, indent_level+1, indent_str)
-		content_foot = html(e.foot, k, i, indent_level+1, indent_str)
-		# FIXME: Fancy pandoc tables, using colspec
-
-	if isinstance(e, TableCell):
-		tag = "td"
-		if e.colspan != 1:
-			attributes += f' colspan="{e.colspan}"'
-		if e.rowspan != 1:
-			attributes += f' rowspan="{e.rowspan}"'
-		aligns = {
-			"AlignLeft": "left",
-			"AlignRight": "right",
-			"AlignCenter": "center"
-		}
-		if e.alignment and e.alignment != "AlignDefault":
-			attributes += f' style="text-align: {aligns[e.alignment]}"'
-
-	# The default which all non-overriding elements get generated by. This
-	# includes elements, which were not explicitly mentioned in this function,
-	# e. g. Strong
-
-	if isinstance(e, Inline):
-		return f'<{tag}{attributes}>{content_head}{html(e.content, k, i, 0, "") if hasattr(e, "_content") else ""}{e.text if hasattr(e, "text") else ""}{content_foot}</{tag}>'
-
-	out_str = ""
-	if not isinstance(e, Plain):
-		out_str += f"{indent_level*indent_str}<{tag}{attributes}>\n"
-	out_str += content_head
-	if hasattr(e, "_content"):
-		if len(e.content) > 0 and isinstance(e.content[0], Inline):
-			out_str += (indent_level+1)*indent_str
-		out_str += html(e.content, k, i, indent_level+1, indent_str)
-	if hasattr(e, "text"):
-		out_str += e.text
-	out_str += f"{content_foot}\n"
-	if not isinstance(e, Plain):
-		out_str += f"{indent_level*indent_str}</{tag}>\n"
-
-	return out_str
-
-

src/formatitko/html_generator.py

@@ -1,4 +1,4 @@
-from panflute import Cite, Emph, Image, LineBreak, Link, Math, Note, RawInline, SmallCaps, Str, Strikeout, Subscript, Superscript, Underline
+from panflute import Cite, Code, Emph, Image, LineBreak, Link, Math, Note, RawInline, SmallCaps, Str, Strikeout, Subscript, Superscript, Underline
 from panflute import BulletList, Citation, CodeBlock, Definition, DefinitionItem, DefinitionList, Header, HorizontalRule, LineBlock, LineItem, ListItem, Null, OrderedList, Para, Plain, RawBlock, TableBody, TableFoot, TableHead
 from panflute import TableRow, TableCell, Caption, Doc
 from panflute import ListContainer, Element
@@ -17,8 +17,10 @@ from .whitespace import NBSP
 from .context import Group, BlockGroup, InlineGroup
 from .output_generator import OutputGenerator
 from .katex import KatexClient
-from .images import ImageProcessor
+from .images import ImageProcessor, ImageProcessorNamespaceSearcher
 from .util import inlinify
+from .elements import FileLink
+
 
 class HTMLGenerator(OutputGenerator):
 	imageProcessor: ImageProcessor
@@ -37,7 +39,7 @@ class HTMLGenerator(OutputGenerator):
 	def escape_special_chars(self, text: str) -> str:
 		text = text.replace("&", "&")
 		text = text.replace("<", "&lt;")
-		text = text.replace(">", "&rt;")
+		text = text.replace(">", "&gt;")
 		text = text.replace("\"", "&quot;")
 		text = text.replace("'", "&#39;")
 		# text = text.replace(" ", "&nbsp;") # Don't replace no-break spaces with HTML escapes, because we trust unicode?
@@ -125,15 +127,26 @@ class HTMLGenerator(OutputGenerator):
 			result = highlight(e.text, lexer, formatter)
 			self.writeraw(result)
 		else:
+			e.text = self.escape_special_chars(e.text)
 			self.generate_simple_tag(e, tag="pre")
 
+	def generate_Code(self, e: Code):
+		e.text = self.escape_special_chars(e.text)
+		self.generate_simple_tag(e)
+
 	def generate_Image(self, e: Image):
 		url = e.url
 
 		additional_args = self.get_image_processor_args(e.attributes)
+		additional_args["context"] = self.context
 
-		# The directory of the current file, will also look for images there.
+		# The directory of the current file relative to the current working directory
 		source_dir = self.context.dir
+		# The directory of the current file relative to the md file we were called on
+		rel_dir = self.context.rel_dir
+
+		searcher = self.imageProcessor.get_searcher_by_path(url, rel_dir, source_dir)
+		url = self.imageProcessor.get_path_without_namespace(url)
 
 		_, ext = os.path.splitext(url)
 		ext = ext[1:]
@@ -143,16 +156,16 @@ class HTMLGenerator(OutputGenerator):
 			# Even supported elements have to be 'converted' because the
 			# processing contains finding and moving them to the output
 			# directory.
-			url = self.imageProcessor.process_image(url, ext, source_dir, **additional_args)
+			url = self.imageProcessor.process_image(url, ext, searcher, **additional_args)
 		elif ext in ["pdf", "epdf","asy"]:
 			# Only relevant for when these were PNGs, leaving this here for future reference.
 			# if not "dpi" in additional_args:
 			#	additional_args["dpi"] = 300
-			url = self.imageProcessor.process_image(url, "svg", source_dir, **additional_args)
+			url = self.imageProcessor.process_image(url, "svg", searcher, **additional_args)
 		elif ext in ["jpg"]:
-			url = self.imageProcessor.process_image(url, "jpeg", source_dir, **additional_args)
+			url = self.imageProcessor.process_image(url, "jpeg", searcher, **additional_args)
 		else:
-			url = self.imageProcessor.process_image(url, "png", source_dir, **additional_args)
+			url = self.imageProcessor.process_image(url, "png", searcher, **additional_args)
 		
 		# Srcset generation - multiple alternative sizes of images browsers can
 		# choose from.
@@ -163,23 +176,27 @@ class HTMLGenerator(OutputGenerator):
 			# This is inspired by @vojta001's blogPhoto shortcode he made for
 			# patek.cz:
 			# https://gitlab.com/patek-devs/patek.cz/-/blob/master/themes/patek/layouts/shortcodes/blogPhoto.html
-			width, height = self.imageProcessor.get_image_size(url, [self.imageProcessor.cache_dir])
+			width, height = self.imageProcessor.get_image_size(searcher.find_image_in_dir(url, searcher.get_cache_dir()))
 			sizes = [(640, 360, 85), (1280, 720, 85), (1920, 1080, 90)] # (widht, height, quality)
 			for size in sizes:
 				if width <= size[0] and height <= size[1]:
-					srcset.append((f'{self.imageProcessor.web_path}/{url}', f'{width}w'))
+					srcset.append((f'{searcher.get_web_path()}/{url}', f'{width}w'))
 					break
 				quality = size[2] if ext == "jpeg" else None
-				cache_img = self.imageProcessor.process_image(url, ext, self.imageProcessor.cache_dir, width=size[0], height=size[1], quality=quality)
-				self.imageProcessor.publish_image(cache_img)
-				srcset.append((f'{self.imageProcessor.web_path}/{cache_img}', f'{size[0]}w'))
+				cache_img = self.imageProcessor.process_image(url, ext, searcher.get_cache_searcher(), width=size[0], height=size[1], quality=quality)
+				searcher.publish_image(cache_img)
+				srcset.append((f'{searcher.get_web_path()}/{cache_img}', f'{size[0]}w'))
 
-		self.imageProcessor.publish_image(url)
-		url = self.imageProcessor.web_path + "/" + url
+		searcher.publish_image(url)
+		url = searcher.get_web_path() + "/" + url
 		
 		attributes = self.common_attributes(e)
 		if "width" in e.attributes:
 			attributes["width"] = e.attributes["width"]
+		if "height" in e.attributes:
+			attributes["height"] = e.attributes["height"]
+		if "title" in e.attributes:
+			attributes["title"] = e.attributes["title"]
 
 		if e.title:
 			attributes["alt"] = e.title
@@ -188,17 +205,38 @@ class HTMLGenerator(OutputGenerator):
 			HTMLGenerator(fake_out, self.katexClient, self.imageProcessor).generate(e.content)
 			attributes["alt"] = fake_out.getvalue()
 		
-		if len(srcset) != 0:
+		if len(srcset) > 1:
 			attributes["src"] = srcset[-1][0]
 			attributes["srcset"] = ", ".join([" ".join(src) for src in srcset])
 		else:
 			attributes["src"] = url
 
+		if e.attributes["no-img-link"]:
+			self.write(self.single_tag("img", attributes))
+			return
+
 		img = RawInline(self.single_tag("img", attributes))
 		link = Link(img, url=url)
 
 		self.generate(link)
 
+	def generate_FileLink(self, e: FileLink):
+		url = e.url
+
+		# The directory of the current file relative to the current working directory
+		source_dir = self.context.dir
+		# The directory of the current file relative to the md file we were called on
+		rel_dir = self.context.rel_dir
+
+		searcher = self.imageProcessor.get_searcher_by_path(url, rel_dir, source_dir)
+		url = self.imageProcessor.get_path_without_namespace(url)
+
+		url = self.imageProcessor.process_image(url, "", searcher, self.context)
+		searcher.publish_image(url)
+		url = searcher.get_web_path() + "/" + url
+
+		self.generate_Link(Link(*e.content, url=url))
+
 	def generate_InlineGroup(self, e: InlineGroup):
 		self.generate_Group(e)
 
@@ -238,7 +276,11 @@ class HTMLGenerator(OutputGenerator):
 			"DisplayMath": True,
 			"InlineMath": False
 		}
-		self.writeln(self.katexClient.render(e.text, {"displayMode": formats[e.format]}))
+		rawhtml = self.katexClient.render(e.text, {"displayMode": formats[e.format]})
+		if (e.format == "InlineMath"):
+			self.write(rawhtml)
+		else:
+			self.writeraw(rawhtml)
 
 	def generate_RawInline(self, e: RawInline):
 		if e.format == "html":
@@ -286,13 +328,13 @@ class HTMLGenerator(OutputGenerator):
 			attributes["style"] = attributes.get("style", "")+f"text-align: {aligns[e.alignment]};"
 		self.generate_simple_tag(e, attributes=attributes)
 
+	def generate_Cite(self, e: Cite):
+		self.generate_simple_tag(e, tag="a", attributes=self.common_attributes(e) | {"href": f"#ref-{e.citations[0].id}"})
+
 	# These are also disabled in pandoc so they shouldn't appear in the AST at all.
 	def generate_Citation(self, e: Citation):
 		self.writeln("<!-- FIXME: Citations not implemented -->")
 
-	def generate_Cite(self, e: Cite):
-		self.generate_simple_tag(e, tag="a", attributes=self.common_attributes(e) | {"href": f"#ref-{e.citations[0].id}"})
-
 	def generate_Definition(self, e: Definition):
 		self.writeln("<!-- FIXME: Definitions not implemented -->")
 
@@ -301,3 +343,34 @@ class HTMLGenerator(OutputGenerator):
 
 	def generate_DefinitionList(self, e: DefinitionList):
 		self.writeln("<!-- FIXME: DefinitionLists not implemented -->")
+
+
+class StandaloneHTMLGenerator(HTMLGenerator):
+	def generate_Doc(self, e: Doc):
+		self.writeraw("<!DOCTYPE html>")
+		self.writeln(self.start_tag("html", attributes={"lang": e.get_metadata("lang", None, True)}))
+		self.writeln(self.start_tag("head"))
+		self.indent_more()
+		self.writeln(self.single_tag("meta", attributes={"charset": "utf-8"}))
+		self.writeln(self.single_tag("meta", attributes={"viewport": "width=device-width, initial-scale=1.0"}))
+		self.writeln(self.single_tag("link", attributes={"rel": "stylesheet", "href": "https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css", "integrity":"sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0", "crossorigin":"anonymous"}))
+		if "title" in e.metadata:
+			self.write(self.start_tag("title"))
+			self.generate(e.metadata["title"])
+			self.write(self.end_tag("title"))
+			self.endln()
+		if "html-head-includes" in e.metadata:
+			self.generate(e.metadata["html-head-includes"])
+		self.indent_less()
+		self.writeln(self.end_tag("head"))
+		self.writeln(self.start_tag("body"))
+		self.indent_more()
+		super().generate_Doc(e)
+		self.indent_less()
+		self.writeln(self.end_tag("body"))
+		self.writeln(self.end_tag("html"))
+
+
+
+
+

src/formatitko/images.py

@@ -4,54 +4,182 @@ import shutil
 import subprocess
 from PIL import Image
 
+from .context import Context
+
+
 class FileInWrongDirError(Exception):
 	pass
 
+
 class ConversionProgramError(Exception):
 	pass
 
+
 class InkscapeError(ConversionProgramError):
 	pass
 
+
 class ImageMagickError(ConversionProgramError):
 	pass
 
+
 class AsyError(ConversionProgramError):
 	pass
 
 
-
-class ImageProcessor:
+class ImageProcessorNamespace:
 	public_dir: str
 	cache_dir: str
 	lookup_dirs: list[str]
 	web_path: str
+	include_src: bool
 
-	def __init__(self, public_dir: str, web_path: str, cache_dir: str, *lookup_dirs: list[str]):
+	def __init__(self, public_dir: str, web_path: str, cache_dir: str, lookup_dirs: list[str], include_src: bool):
 		self.public_dir = public_dir
 		self.cache_dir = cache_dir
 		self.lookup_dirs = lookup_dirs
 		self.web_path = web_path if web_path[-1] != "/" else web_path[:-1]
-		if not os.path.exists(self.public_dir):
-			os.mkdir(self.public_dir)
-		if not os.path.exists(self.cache_dir):
-			os.mkdir(self.cache_dir)
+		self.include_src = include_src
 
-	def process_image(self, input_filename: str, format: str, source_dir: str, width: int=None, height:int=None, quality: int=None, dpi: int=None, fit: bool=True, deps: list[str]=[]) -> str:
+class ImageProcessorSearcher:
+	def get_lookup_dirs(self) -> list[str]:
+		return []
+
+	def get_cache_dir(self) -> str:
+		return ""
+
+	def get_public_dir(self) -> str:
+		return ""
+
+	def get_web_path(self) -> str:
+		return ""
+	
+	def find_image_in_dir(self, input_filename: str, dir: str) -> Union[str, None]:
+		if os.path.isfile(dir + "/" + input_filename):
+			return dir + "/" + input_filename
+		else:
+			return None
+
+	def find_image(self, input_filename: str) -> Union[str, None]:
+		for dir in self.get_lookup_dirs():
+			image = self.find_image_in_dir(input_filename, dir)
+			if image:
+				return image
+		return None
+
+	def publish_image(self, target_name, relative: bool=True) -> str:
+		cache_path = self.get_cache_dir() + "/" + target_name
+		if not os.path.isfile(cache_path):
+			raise FileNotFoundError(f'Image {target_name} not cached')
+		target_path = self.get_public_dir() + "/" + target_name
+		try:
+			if os.path.exists(target_path):
+				if os.path.getmtime(cache_path) > os.path.getmtime(target_path):
+					os.remove(target_path)
+					os.link(cache_path, target_path)
+			else:
+				os.link(cache_path, target_path)
+		except OSError as e:
+			if e.errno == 18: # Invalid cross-device link: cache and public dirs are on different devices, don't hardlink, copy
+				shutil.copyfile(cache_path, target_path)
+			else:
+				raise e
+		return target_name if relative else target_path
+
+
+
+class ImageProcessorCacheSearcher(ImageProcessorSearcher):
+	cache_dir: str
+
+	def __init__(self, cache_dir: str):
+		self.cache_dir = cache_dir
+		if not os.path.exists(self.cache_dir):
+			os.makedirs(self.cache_dir, exist_ok=True)
+
+	def get_lookup_dirs(self) -> list[str]:
+		return [self.cache_dir]
+
+	def get_cache_dir(self) -> str:
+		return self.cache_dir
+
+	def get_public_dir(self) -> str:
+		return ""
+
+	def get_web_path(self) -> str:
+		return ""
+
+	def publish_image(self, target_name, relative: bool=True) -> str:
+		raise NotImplementedError();
+
+class ImageProcessorNamespaceSearcher(ImageProcessorSearcher):
+	namespace: ImageProcessorNamespace
+	rel_dir: str
+	source_dir: str
+
+	def __init__(self, namespace: ImageProcessorNamespace, rel_dir: str, source_dir: str):
+		self.namespace = namespace
+		self.rel_dir = rel_dir
+		self.source_dir = source_dir
+
+	def get_lookup_dirs(self) -> list[str]:
+		return self.namespace.lookup_dirs + ([self.source_dir] if self.namespace.include_src else [])
+
+	def transform_path(self, path: str) -> str:
+		return path.replace("$dir", self.rel_dir)
+
+	def get_cache_dir(self) -> str:
+		cache_dir = self.transform_path(self.namespace.cache_dir)
+		if not os.path.exists(cache_dir):
+			os.makedirs(cache_dir, exist_ok=True)
+		return cache_dir
+
+	def get_public_dir(self) -> str:
+		public_dir = self.transform_path(self.namespace.public_dir)
+		if not os.path.exists(public_dir):
+			os.makedirs(public_dir, exist_ok=True)
+		return public_dir
+
+	def get_web_path(self) -> str:
+		return self.transform_path(self.namespace.web_path)
+
+	def get_cache_searcher(self) -> ImageProcessorCacheSearcher:
+		return ImageProcessorCacheSearcher(self.get_cache_dir())
+
+class ImageProcessor:
+	namespaces: dict[str, ImageProcessorNamespace]
+
+	def __init__(self, namespaces: dict[str, ImageProcessorNamespace]):
+		self.namespaces = namespaces
+
+	def get_namespace_by_path(self, path: str) -> ImageProcessorNamespace:
+		return self.namespaces[path.split(":")[0] if ":" in path else ""]
+
+	def get_path_without_namespace(self, path: str) -> str:
+		if len(path.split(":")) <= 1:
+			return path
+		return ":".join(path.split(":")[1:])
+
+	def get_searcher_by_path(self, path: str, rel_dir: str, source_dir: str) -> ImageProcessorNamespaceSearcher:
+		return ImageProcessorNamespaceSearcher(self.get_namespace_by_path(path), rel_dir, source_dir)
+
+	def process_image(self, input_filename: str, format: str, searcher: ImageProcessorSearcher, context: Context=None, width: int=None, height:int=None, quality: int=None, dpi: int=None, fit: bool=True, deps: list[str]=[]) -> str:
 		name = os.path.basename(input_filename)
 		base, ext = os.path.splitext(name)
 		ext = ext[1:]
-		full_path = self.find_image(input_filename, [source_dir])
+		full_path = searcher.find_image(input_filename)
 		if full_path is None:
-			raise FileNotFoundError(f'Image {input_filename} not found in {self.lookup_dirs} or {source_dir}.')
+			raise FileNotFoundError(f'Image {input_filename} not found in {searcher.get_lookup_dirs()}.')
 
 		if format == "jpg":
 			format = "jpeg"
 
+		if format == "":
+			format = ext
+
 		# Locate all dependencies
 		deps_full = [full_path]
 		for dep in deps:
-			dep_full_path = self.find_image(dep, [source_dir])
+			dep_full_path = searcher.find_image(dep)
 			if dep_full_path is None:
 				raise FileNotFoundError(f'Image dependency {dep} not found.')
 			deps_full.append(dep_full_path)
@@ -65,7 +193,7 @@ class ImageProcessor:
 		if quality is not None:
 			suffix += f'_q{quality}'
 		target_name = base+suffix+"."+format
-		target_path = self.cache_dir + "/" + target_name
+		target_path = searcher.get_cache_dir() + "/" + target_name
 	
 		# Only regenerate if the file doesn't already exist and no dependencies are newer
 		if not os.path.isfile(target_path) or self.is_outdated(target_path, deps_full):
@@ -80,13 +208,13 @@ class ImageProcessor:
 
 			# Try to find the converted filename in lookup_dirs, if you find
 			# it, don't convert, just copy.
-			elif self.find_image(target_name, [source_dir]) is not None and not self.is_outdated(self.find_image(target_name, [source_dir]), deps):
-				shutil.copyfile(self.find_image(target_name, [source_dir]), target_path)
+			elif searcher.find_image(target_name) is not None and not self.is_outdated(searcher.find_image(target_name), deps):
+				shutil.copyfile(searcher.find_image(target_name), target_path)
 
 			# Process asymptote
 			elif ext == "asy":
 				# Collect dependencies
-				deps_dir = self.cache_dir + "/" + name + "_deps"
+				deps_dir = searcher.get_cache_dir() + "/" + name + "_deps"
 				if not os.path.isdir(deps_dir):
 					os.mkdir(deps_dir)
 				for dep_full in deps_full:
@@ -96,7 +224,7 @@ class ImageProcessor:
 				dpi_arg = ['-render', str(dpi/72)] if dpi is not None else []
 				if subprocess.run(['asy', name, '-o', target_name, '-f', format, *dpi_arg], cwd=deps_dir).returncode != 0:
 					raise AsyError(f"Could not convert '{full_path}' to '{format}'")
-				shutil.move(deps_dir + "/" + target_name, self.cache_dir + "/" + target_name)
+				shutil.move(deps_dir + "/" + target_name, searcher.get_cache_dir() + "/" + target_name)
 
 			# Convert SVGs using inkscape
 			elif ext == "svg":
@@ -114,6 +242,8 @@ class ImageProcessor:
 				if subprocess.run(['convert', *density_arg, full_path, *resize_arg, *quality_arg, target_path]).returncode != 0:
 					raise ImageMagickError(f"Could not convert '{full_path}' to '{format}'")
 
+		if context is not None:
+			context.add_deps(deps_full)
 		return target_name
 
 	def is_outdated(self, target: str, deps: list[str]):
@@ -124,38 +254,7 @@ class ImageProcessor:
 				return True
 		return False
 
-	def publish_image(self, target_name, relative: bool=True) -> str:
-		import sys
-		cache_path = self.cache_dir + "/" + target_name
-		if not os.path.isfile(cache_path):
-			raise FileNotFoundError(f'Image {target_name} not cached')
-		target_path = self.public_dir + "/" + target_name
-		try:
-			if os.path.exists(target_path):
-				if os.path.getmtime(cache_path) > os.path.getmtime(target_path):
-					os.remove(target_path)
-					os.link(cache_path, target_path)
-			else:
-				os.link(cache_path, target_path)
-		except OSError as e:
-			if e.errno == 18: # Invalid cross-device link: cache and public dirs are on different devices, don't hardlink, copy
-				shutil.copyfile(cache_path, target_path)
-			else:
-				raise e
-		return target_name if relative else target_path
-
-	def get_image_size(self, input_filename: str, additional_dirs: list[str]=[]) -> tuple[int, int]:
-		full_path = self.find_image(input_filename, additional_dirs)
-		if full_path is None:
-			raise FileNotFoundError(f'Image {input_filename} not found.')
+	def get_image_size(self, full_path: str) -> tuple[int, int]:
 		# Getting image size using ImageMagick is slow. VERY
 		return Image.open(full_path).size
 
-
-	def find_image(self, input_filename: str, additional_dirs: list[str]=[]) -> Union[str, None]:
-		for dir in [*self.lookup_dirs, *additional_dirs]:
-			if os.path.isfile(dir + "/" + input_filename):
-				return dir + "/" + input_filename
-	
-
-

src/formatitko/katex-server

@@ -0,0 +1 @@
+Subproject commit 211cb2010e23265be599819c5f79f66f0abd62d1

src/formatitko/katex-server/.gitignore

@@ -1 +0,0 @@
-node_modules

src/formatitko/katex-server/README.md

@@ -1 +0,0 @@
-This was made by Standa Lukeš @exyi

src/formatitko/katex-server/index.js

@@ -1 +0,0 @@
-console.log(require('katex').renderToString('\\frac{2a}{b}'))

src/formatitko/katex-server/index.mjs

@@ -1,131 +0,0 @@
-// KaTeX rendering server
-// Listens on unix socket, path is provided as first argument
-// Expects JSON lines, each line is a query with the following schema:
-// {
-//   formulas: [
-//     {
-//       tex: string,
-//       options?: object
-//     }
-//   ],
-//   options?: object
-// }
-
-// see https://katex.org/docs/options.html for list of available options
-// If options formulas[].options field is used, the global options field is ignored.
-
-// For each line, returns one JSON line with the following schema:
-// {
-//   results: [
-//     { html?: string } | { error?: string }
-//   ]
-// } | { error?: string }
-
-
-// If one formula is invalid, the error in results is used
-// If the entire query is invalid (couldn't parse JSON, for example), the outer error field is used
-
-
-import katex from 'katex'
-import net from 'net'
-import * as readline from 'readline'
-
-const myArgs = process.argv.slice(2)
-
-const unixSocketPath = myArgs[0]
-if (!unixSocketPath) {
-    console.error('you must specify socket path')
-    process.exit(1)
-}
-
-// This server listens on a Unix socket at /var/run/mysocket
-var unixServer = net.createServer(handleClient);
-unixServer.listen(unixSocketPath);
-console.log("OK")
-
-function handleExit(signal) {
-    // unixServer.emit('close')
-    unixServer.close(function () {
-
-    });
-    process.exit(0); // put this into the callback to avoid closing open connections
-}
-process.on('SIGINT', handleExit);
-process.on('SIGQUIT', handleExit);
-process.on('SIGTERM', handleExit);
-process.on('exit', handleExit);
-
-const defaultOptions = {}
-
-/**
- * @param {net.Socket} socket
- * @returns {Promise<void>}
- * */
-function socketWrite(socket, data) {
-    return new Promise((resolve, reject) => {
-        socket.write(data, (err) => {
-            if (err) {
-                reject(err)
-            } else {
-                resolve()
-            }
-        })
-    })
-}
-
-/**
- * @param {net.Socket} client
- * */
-async function handleClient(client) {
-    const rl = readline.createInterface({ input: client })
-    
-    /* Added by GS: A stack of katex's `macros` objects, each group inherits
-     * the one from the parent group and can add its own stuff without
-     * affecting the parent.
-     */
-    let macroStack = [{}]
-    for await (const line of rl) {
-        try {
-            // The custom commands for pushing and popping the macro stack.
-            if (line === "begingroup") {
-                // Copy the current state of macros and push it onto the stack.
-                macroStack.push({...macroStack.slice(-1)[0]})
-                continue
-            } else if (line === "endgroup") {
-                macroStack.pop()
-                continue
-            } else if (line === "init") {
-                macroStack = [{}]
-                continue
-            }
-            const query = JSON.parse(line)
-            const results = []
-            for (const input of query.formulas) {
-                const options = input.options ?? query.options ?? defaultOptions
-                // Add macros from the macros option
-                if (options.macros) {
-                    for (const macro of Object.keys(options.macros)) {
-                        macroStack.slice(-1)[macro] = options.macros[macro]
-                    }
-                }
-                options.macros = macroStack.slice(-1)[0]
-                // Enforce globalGroup option, katex then saves created macros
-                // into the options.macros object.
-                options.globalGroup = true
-                try {
-                    const html = katex.renderToString(input.tex, options)
-                    results.push({ html })
-                } catch (e) {
-                    results.push({ error: String(e) })
-                }
-            }
-            await socketWrite(client, JSON.stringify({ results }, null, query.debug ? '  ' : undefined))
-            await socketWrite(client, '\n')
-        } catch (e) {
-            console.error(e)
-            await socketWrite(client, JSON.stringify({ error: String(e) }))
-            await socketWrite(client, '\n')
-        }
-    }
-}
-

src/formatitko/katex-server/package-lock.json

@@ -1,39 +0,0 @@
-{
-  "name": "ksp-katex-server",
-  "version": "1.0.0",
-  "lockfileVersion": 3,
-  "requires": true,
-  "packages": {
-    "": {
-      "name": "ksp-katex-server",
-      "version": "1.0.0",
-      "license": "ISC",
-      "dependencies": {
-        "katex": "^0.16.3"
-      }
-    },
-    "node_modules/commander": {
-      "version": "8.3.0",
-      "resolved": "https://registry.npmjs.org/commander/-/commander-8.3.0.tgz",
-      "integrity": "sha512-OkTL9umf+He2DZkUq8f8J9of7yL6RJKI24dVITBmNfZBmri9zYZQrKkuXiKhyfPSu8tUhnVBB1iKXevvnlR4Ww==",
-      "engines": {
-        "node": ">= 12"
-      }
-    },
-    "node_modules/katex": {
-      "version": "0.16.3",
-      "resolved": "https://registry.npmjs.org/katex/-/katex-0.16.3.tgz",
-      "integrity": "sha512-3EykQddareoRmbtNiNEDgl3IGjryyrp2eg/25fHDEnlHymIDi33bptkMv6K4EOC2LZCybLW/ZkEo6Le+EM9pmA==",
-      "funding": [
-        "https://opencollective.com/katex",
-        "https://github.com/sponsors/katex"
-      ],
-      "dependencies": {
-        "commander": "^8.0.0"
-      },
-      "bin": {
-        "katex": "cli.js"
-      }
-    }
-  }
-}

src/formatitko/katex-server/package.json

@@ -1,14 +0,0 @@
-{
-  "name": "ksp-katex-server",
-  "version": "1.0.0",
-  "description": "",
-  "main": "index.mjs",
-  "scripts": {
-    "test": "echo \"Error: no test specified\" && exit 1"
-  },
-  "author": "",
-  "license": "ISC",
-  "dependencies": {
-    "katex": "^0.16.3"
-  }
-}

src/formatitko/katex.py

@@ -3,6 +3,7 @@ import subprocess
 import tempfile
 import json
 import os
+import shutil
 
 class KatexError(Exception):
 	pass
@@ -20,8 +21,10 @@ class KatexClient:
 	_socket_file: str
 	_temp_dir: tempfile.TemporaryDirectory[str]
 	_connected: bool
+	_katex_server_path: str
 
-	def __init__(self, socket: str=None, connect: bool=True):
+	def __init__(self, socket: str=None, connect: bool=True, katex_server_path: str=None):
+		self._katex_server_path = katex_server_path
 		if socket is not None:
 			self._socket_file = socket
 		else:
@@ -38,20 +41,21 @@ class KatexClient:
 		self._temp_dir = tempfile.TemporaryDirectory(prefix='formatitko')
 		self._socket_file = self._temp_dir.name + "/katex-socket"
 
-		srcdir = os.path.dirname(os.path.realpath(__file__))
+		if self._katex_server_path is None:
+			
+			srcdir = os.path.dirname(os.path.realpath(__file__))
 
-		# Test if `node_modules` directory exists and if not, run `npm install`
-		if not os.path.isdir(srcdir + "/katex-server/node_modules"):
-			print("Installing node dependencies for the first time...")
-			try:
-				subprocess.run(["npm", "install"], cwd=srcdir+"/katex-server", check=True)
-			except subprocess.CalledProcessError as e:
-				if e.returncode == 127:
+			# Test if `node_modules` directory exists and if not, run `npm install`
+			if not os.path.isdir(srcdir + "/katex-server/node_modules"):
+				print("Installing node dependencies for the first time...")
+				npm = shutil.which("npm") or shutil.which("yarnpkg")
+				if npm is None:
 					raise NPMNotFoundError("npm not found. Node.js is required to use KaTeX.")
-				else:
-					raise e
+				subprocess.run([npm, "install"], cwd=srcdir+"/katex-server", check=True)
+			
+			self._katex_server_path = srcdir + "/katex-server/index.mjs"
 
-		self._server_process = subprocess.Popen(["node", srcdir + "/katex-server/index.mjs", self._socket_file], stdout=subprocess.PIPE)
+		self._server_process = subprocess.Popen(["node", self._katex_server_path, self._socket_file], stdout=subprocess.PIPE)
 
 		ok = self._server_process.stdout.readline()
 		if ok != b"OK\n":
@@ -79,7 +83,7 @@ class KatexClient:
 		if "error" in response:
 			raise KatexServerError(response["error"])
 		if "error" in response["results"][0]:
-			raise KatexError(response["results"][0]["error"])
+			raise KatexError(response["results"][0]["error"] + " in $" + tex + "$")
 		else:
 			return response["results"][0]["html"]
 

src/formatitko/nop_processor.py

@@ -6,16 +6,21 @@ from panflute import MetaValue
 from typing import Union, Callable
 
 from .whitespace import NBSP
-from .elements import FQuoted
-from .context import Group, InlineGroup, BlockGroup
+from .elements import FQuoted, FileLink
+from .context import Group, InlineGroup, BlockGroup, Context
 from .whitespace import Whitespace
 from .command import BlockCommand, InlineCommand, CodeCommand, Command
+from .output_generator import FormatitkoRecursiveError
 
 ELCl = Union[Element, ListContainer, list[Union[Element, ListContainer]]]
 
+class DoubleDocError(Exception):
+	"TransformProcessor should only ever see a single Doc."
+	pass
 
 class NOPProcessor:
 	TYPE_DICT: dict[type, Callable]
+	context: Union[Context, None] = None
 
 	class UnknownElementError(Exception):
 		f"An unknown Element has been passed to the NOPProcessor, probably because panflute introduced a new one."
@@ -83,6 +88,7 @@ class NOPProcessor:
 			Underline: self.transform_Underline,
 			NBSP: self.transform_NBSP,
 			FQuoted: self.transform_FQuoted,
+			FileLink: self.transform_FileLink,
 
 			InlineCommand: self.transform_InlineCommand,
 			BlockCommand: self.transform_BlockCommand,
@@ -96,32 +102,45 @@ class NOPProcessor:
 		return []
 
 	def transform(self, e: ELCl) -> ELCl:
-		if isinstance(e, list):
-			return self.transform_list(e)
-		elif isinstance(e, ListContainer):
-			return self.transform_ListContainer(e)
-
-		for transformer in self.get_pretransformers():
-			e = transformer(e)
-
 		try:
-			e = self.TYPE_DICT[type(e)](e)
-		except KeyError:
-			raise self.UnknownElementError(type(e))
+			if isinstance(e, list):
+				return self.transform_list(e)
+			elif isinstance(e, ListContainer):
+				return self.transform_ListContainer(e)
 
-		for transformer in self.get_posttransformers():
-			e = transformer(e)
+			for transformer in self.get_pretransformers():
+				e = transformer(e)
 
-		return e
+			try:
+				method = self.TYPE_DICT[type(e)]
+			except KeyError:
+				raise self.UnknownElementError(type(e))
+
+			e = method(e)
+
+			for transformer in self.get_posttransformers():
+				e = transformer(e)
+
+			return e
+		except FormatitkoRecursiveError as err:
+			if not isinstance(e, ListContainer):
+				err.add_element(e)
+			raise err
+		except Exception as err:
+			raise FormatitkoRecursiveError(e, self.context) from err
 
 	def transform_list(self, e: list[Union[Element, ListContainer]]) -> list[Union[Element, ListContainer]]:
-		for i in range(len(e)):
+		i = 0
+		while i < len(e): # The length of the list can change mid-transformation, so we need to check the length each time
 			e[i] = self.transform(e[i])
+			i-=-1
 		return e
 
 	def transform_ListContainer(self, e: ListContainer) -> ListContainer:
-		for i in range(len(e)):
+		i = 0
+		while i < len(e): # The length of the list can change mid-transformation, so we need to check the length each time
 			e[i] = self.transform(e[i])
+			i-=-1
 		return e
 
 
@@ -281,6 +300,10 @@ class NOPProcessor:
 		e.content = self.transform(e.content)
 		return e
 
+	def transform_FileLink(self, e: FileLink) -> FileLink:
+		e.content = self.transform(e.content)
+		return e
+
 	def transform_Figure(self, e: Figure) -> Figure:
 		e.content = self.transform(e.content)
 		e.caption = self.transform(e.caption)
@@ -293,6 +316,9 @@ class NOPProcessor:
 		return e
 
 	def transform_Doc(self, e: Doc) -> Doc:
+		if self.context is not None:
+			raise DoubleDocError()
+		self.context = Context(e, self.root_file_path)
 		e.content = self.transform(e.content)
 		return e
 

src/formatitko/output_generator.py

@@ -3,20 +3,58 @@ from panflute import Cite, Code, Emph, Image, LineBreak, Link, Math, Note, Quote
 from panflute import BlockQuote, BulletList, Citation, CodeBlock, Definition, DefinitionItem, DefinitionList, Div, Figure, Header, HorizontalRule, LineBlock, LineItem, ListItem, MetaBlocks, MetaBool, MetaInlines, MetaList, MetaMap, MetaString, Null, OrderedList, Para, Plain, RawBlock, Table, TableBody, TableFoot, TableHead
 from panflute import TableRow, TableCell, Caption, Doc
 from panflute import MetaValue
+from panflute import stringify
 from typing import Union, Callable
 
 from .whitespace import NBSP
-from .elements import FQuoted
+from .elements import FQuoted, FileLink
 from .context import Group, InlineGroup, BlockGroup, Context
 
-import re
+
+import sys
 
 class UnknownElementError(Exception):
 	"An unknown Element has been passed to the OutputGenerator, probably because panflute introduced a new one."
 	pass
 
+
+class FormatitkoRecursiveError(Exception):
+	"A generic exception which wraps other exceptions and adds element-based traceback"
+	elements: list[Union[Element, ListContainer, list[Union[Element, ListContainer]]]]
+	context: Context
+
+	def __init__(self, e: Union[Element, ListContainer, list[Union[Element, ListContainer]]], context: Context, *args):
+		self.elements = [e]
+		self.context = context
+		super().__init__(args)
+
+	def add_element(self, e: Union[Element, ListContainer, list[Union[Element, ListContainer]]]):
+		self.elements.append(e)
+
+	def pretty_print(self, tracebacklimit: int=0):
+		def eprint(*args, **kwargs):
+			print(*args, file=sys.stderr, **kwargs)
+
+		def print_filename_recursive(context: Context):
+			return context.path +\
+				((" (included from " + print_filename_recursive(context.parent) + ")") if context.parent else "")
+		eprint(f"Error occured in file {print_filename_recursive(self.context)} in ", end="")
+		line = None
+		for i in range(len(self.elements)-1, 0, -1):
+			if hasattr(self.elements[i], "content") and len(self.elements[i].content) > 0 and isinstance(self.elements[i].content[0], Inline) and line is None:
+				line = self.elements[i]
+			eprint(type(self.elements[i]).__name__ + "[" + (str(self.elements[i-1].index) if isinstance(self.elements[i-1].index, int) else "") + "]", end=": ")
+		if line:
+			eprint()
+			eprint('on line: "' + stringify(line).strip() + '"', end="")
+		eprint()
+		eprint("in element: " + str(self.elements[0]).replace("\n", "\\n"))
+		sys.tracebacklimit = tracebacklimit
+		raise self.__cause__ from None
+
+
 class OutputGenerator:
-	_at_start_of_line: bool
+	_empty_lines: int
 	context: Union[Context, None]
 	indent_level: int
 	indent_str: str
@@ -29,7 +67,7 @@ class OutputGenerator:
 		self.output_file = output_file
 		self.indent_str = indent_str
 		self.indent_level = initial_indent_level
-		self._at_start_of_line = True
+		self._empty_lines = 1
 		self.context = None
 
 		self.TYPE_DICT_MISC = {
@@ -89,6 +127,7 @@ class OutputGenerator:
 			Underline: self.generate_Underline,
 			NBSP: self.generate_NBSP,
 			FQuoted: self.generate_FQuoted,
+			FileLink: self.generate_FileLink,
 			InlineGroup: self.generate_InlineGroup
 		}
 
@@ -101,28 +140,36 @@ class OutputGenerator:
 		}
 
 	def generate(self, e: Union[Element, ListContainer, list[Union[Element, ListContainer]]]):
-		if isinstance(e, Group):
-			old_context = self.context
-			self.context = e.context
-		if isinstance(e, list):
-			self.generate_list(e)
-		elif isinstance(e, ListContainer):
-			self.generate_ListContainer(e)
-		elif isinstance(e, Inline):
-			self.generate_Inline(e)
-		elif isinstance(e, Block):
-			self.generate_Block(e)
-		elif isinstance(e, MetaValue):
-			self.generate_MetaValue(e)
-		elif isinstance(e, MetaList):
-			self.generate_MetaList(e)
-		else:
-			try:
-				self.TYPE_DICT_MISC[type(e)](e)
-			except KeyError:
-				raise UnknownElementError(type(e))
-		if isinstance(e, Group):
-			self.context = old_context
+		try:
+			if isinstance(e, Group):
+				old_context = self.context
+				self.context = e.context
+			if isinstance(e, list):
+				self.generate_list(e)
+			elif isinstance(e, ListContainer):
+				self.generate_ListContainer(e)
+			elif isinstance(e, Inline):
+				self.generate_Inline(e)
+			elif isinstance(e, Block):
+				self.generate_Block(e)
+			elif isinstance(e, MetaValue):
+				self.generate_MetaValue(e)
+			elif isinstance(e, MetaList):
+				self.generate_MetaList(e)
+			else:
+				try:
+					method = self.TYPE_DICT_MISC[type(e)]
+				except KeyError as err:
+					raise UnknownElementError(type(e)) from err
+				method(e)
+			if isinstance(e, Group):
+				self.context = old_context
+		except FormatitkoRecursiveError as err:
+			if not isinstance(e, ListContainer):
+				err.add_element(e)
+			raise err
+		except Exception as err:
+			raise FormatitkoRecursiveError(e, self.context) from err
 
 	def escape_special_chars(self, text: str) -> str:
 		return text
@@ -136,29 +183,32 @@ class OutputGenerator:
 	def indent_less(self):
 		self.indent_level -= 1
 
-	def write(self, text: str):
-		if self._at_start_of_line:
+	def write(self, text: str=""):
+		if self._empty_lines > 0:
 			self.output_file.write(self.indent())
 		self.output_file.write(text)
-		self._at_start_of_line = False
+		self._empty_lines = 0
 
-	def writeln(self, text: str):
-		if not self._at_start_of_line:
+	def writeln(self, text: str=""):
+		if self._empty_lines == 0:
 			self.output_file.write("\n")
 		self.output_file.write(self.indent())
 		self.output_file.write(text+"\n")
-		self._at_start_of_line = True
+		self._empty_lines = 1
 	
-	def writeraw(self, text: str):
-		if not self._at_start_of_line:
+	def writeraw(self, text: str=""):
+		if self._empty_lines == 0:
 			self.output_file.write("\n")
 		self.output_file.write(text+"\n")
-		self._at_start_of_line = True
+		self._empty_lines = 1
 	
-	def endln(self):
-		if not self._at_start_of_line:
+	def ensure_empty(self, n: int=1):
+		while self._empty_lines < n:
 			self.output_file.write("\n")
-		self._at_start_of_line = True
+			self._empty_lines+=1
+
+	def endln(self):
+		self.ensure_empty(1)
 
 	def start_tag(self, tag: str, attributes: dict[str,str]={}) -> str:
 		return tag
@@ -223,9 +273,9 @@ class OutputGenerator:
 		self.write(self.end_tag(tag))
 
 	def generate_raw_block_tag(self, tag: str, text: str, attributes: dict[str,str]={}):
-		self.writeln(self.start_tag(tag, attributes))
+		self.writeraw(self.start_tag(tag, attributes))
 		self.writeraw(text)
-		self.writeln(self.end_tag(tag))
+		self.writeraw(self.end_tag(tag))
 
 	def generate_empty_block_tag(self, tag: str, attributes: dict[str,str]={}):
 		self.writeln(self.single_tag(tag, attributes))
@@ -244,9 +294,10 @@ class OutputGenerator:
 
 	def generate_MetaValue(self, e: MetaValue):
 		try:
-			self.TYPE_DICT_META[type(e)](e)
+			method = self.TYPE_DICT_META[type(e)]
 		except KeyError:
 			self.generate(e.content)
+		method(e)
 
 	def generate_MetaBlocks(self, e: MetaBlocks):
 		self.generate(e.content)
@@ -255,16 +306,23 @@ class OutputGenerator:
 		self.generate(e.content)
 
 	def generate_MetaBool(self, e: MetaBool):
-		self.generate_simple_tag(e)
+		if e.boolean:
+			self.write("True")
+		else:
+			self.write("False")
 
 	def generate_MetaMap(self, e: MetaMap):
 		self.generate_simple_tag(e)
 
 	def generate_MetaString(self, e: MetaString):
-		self.generate_simple_tag(e)
+		self.write(e.text)
 
 	def generate_Inline(self, e: Inline):
-		self.TYPE_DICT_INLINE[type(e)](e)
+		try:
+			method = self.TYPE_DICT_INLINE[type(e)]
+		except KeyError as err:
+			raise UnknownElementError(type(e)) from err
+		method(e)
 
 	def generate_Str(self, e: Str):
 		self.write(self.escape_special_chars(e.text))
@@ -310,7 +368,9 @@ class OutputGenerator:
 				self.write("\"")
 				self.generate(e.content)
 				self.write("\"")
-
+	
+	def generate_FileLink(self, e: FileLink):
+		self.generate_simple_tag(e)
 
 	# Inline Elements
 	def generate_Cite(self, e: Cite):
@@ -365,7 +425,11 @@ class OutputGenerator:
 
 
 	def generate_Block(self, e: Block):
-		self.TYPE_DICT_BLOCK[type(e)](e)
+		try:
+			method = self.TYPE_DICT_BLOCK[type(e)]
+		except KeyError as err:
+			raise UnknownElementError(type(e)) from err
+		method(e)
 
 
 	# Block elements
@@ -430,12 +494,14 @@ class OutputGenerator:
 		self.generate_simple_tag(e)
 
 	def generate_Doc(self, e: Doc):
+		if "header-includes" in e.metadata:  # This is the pandoc way of doing things
+			self.generate(e.metadata["header-includes"])
 		if "header_content" in e.metadata:
 			self.generate(e.metadata["header_content"])
 		self.generate_simple_tag(e)
 		if "footer_content" in e.metadata:
 			self.generate(e.metadata["footer_content"])
-	
+
 	def generate_BlockGroup(self, e: BlockGroup):
 		self.generate_simple_tag(e)
 

src/formatitko/tex.py

@@ -1,270 +0,0 @@
-from panflute import *
-import os
-from typing import Union
-
-from .whitespace import NBSP
-from .elements import FQuoted
-from .util import inlinify
-from .context import Group
-from .images import ImageProcessor
-
-# Heavily inspired by: git://git.ucw.cz/labsconf2022.git
-def tex(e: Union[Element, ListContainer], i: ImageProcessor, indent_level: int=0, indent_str: str="\t") -> str:
-	
-	# `only` attribute which makes transformed elements appear only in tex
-	# output or html output
-	if hasattr(e, "attributes") and "only" in e.attributes and e.attributes["only"] != "tex":
-		return ""
-
-	if isinstance(e, ListContainer):
-		return ''.join([tex(child, i, indent_level, indent_str) for child in e])
-
-	# Bits from which the final element output is built at the end of this
-	# function. Most elements override this by returning their own output.
-	content_foot = ""
-	content_head = ""
-
-	arguments = ""
-	open = "{"
-	close = "}"
-
-	tag = e.tag.lower()
-
-	tags = {
-		Header: "h"+chr(64 + e.level) if isinstance(e, Header) else "",
-	}
-	if type(e) in tags:
-		tag = tags[type(e)]
-
-	# These are also disabled in pandoc so they shouldn't appear in the AST at all.
-	not_implemented = {
-		Citation: True,
-		Cite: True,
-		Definition: True,
-		DefinitionItem: True,
-		DefinitionList: True
-	}
-	if type(e) in not_implemented:
-		return f'% FIXME: {type(e)}s not implemented \n'
-	
-	# Elements which can be represented by a simple string
-	simple_string = {
-		NBSP: "~",
-		Space: " ",
-		Null: "",
-		LineBreak: f"\\\\",
-		SoftBreak: f" ",
-		HorizontalRule: "\\hr\n\n"
-	}
-	if type(e) in simple_string:
-		return simple_string[type(e)]
-
-	# Simplest basic elements
-	if isinstance(e, Str):
-		return e.text.replace(" ", "~")
-
-	if isinstance(e, Para):
-		return tex(e.content, i, 0, "")+"\n\n"
-
-	if isinstance(e, Span) or isinstance(e, Plain):
-		return tex(e.content, i, 0, "")
-
-	# Overriding elements with their own returns
-	if isinstance(e, Image):
-		url = e.url
-
-		# TODO: This should use OutputGenerator's get_image_processor_args
-		# Attributes → image processor args
-		additional_args = {}
-		if "file-width" in e.attributes:
-			additional_args["width"] = int(e.attributes["file-width"])
-		if "file-height" in e.attributes:
-			additional_args["height"] = int(e.attributes["file-height"])
-		if "file-quality" in e.attributes:
-			additional_args["quality"] = int(e.attributes["file-quality"])
-		if "file-dpi" in e.attributes:
-			additional_args["dpi"] = int(e.attributes["file-dpi"])
-
-		# The directory of the current file, will also look for images there.
-		source_dir = e.attributes["source_dir"]
-
-		_, ext = os.path.splitext(url)
-		ext = ext[1:]
-
-		# Conversions between various formats.
-		if ext in ["pdf", "png", "jpeg"]:
-			# Even supported elements have to be 'converted' because the
-			# processing contains finding and moving them to the cache
-			# directory.
-			url = i.process_image(url, ext, source_dir, **additional_args)
-		elif ext in ["svg"]:
-			url = i.process_image(url, "pdf", source_dir, **additional_args)
-		elif ext in ["epdf"]:
-			url = i.process_image(url, "pdf", source_dir, **additional_args)
-		elif ext in ["jpg"]:
-			url = i.process_image(url, "jpeg", source_dir, **additional_args)
-		else:
-			url = i.process_image(url, "pdf", source_dir, **additional_args)
-
-		url = i.find_image(url, [i.cache_dir])
-		width = ""
-		if "width" in e.attributes:
-			width = e.attributes["width"]
-			# 50% → 0.5\hsize
-			if e.attributes["width"][-1] == "%":
-				width = str(int(e.attributes["width"][:-1])/100) + "\\hsize"
-			width = "width " + width
-		return f'\\image{{{width}}}{{{url}}}'
-
-	if isinstance(e, FQuoted):
-		if e.style == "cs":
-			if e.quote_type == "SingleQuote":
-				return f'‚{tex(e.content, i, 0, "")}‘'
-			elif e.quote_type == "DoubleQuote":
-				return f'„{tex(e.content, i, 0, "")}“'
-		elif e.style == "en":
-			if e.quote_type == "SingleQuote":
-				return f'‘{tex(e.content, i, 0, "")}’'
-			elif e.quote_type == "DoubleQuote":
-				return f'“{tex(e.content, i, 0, "")}”'
-		else:
-			if e.quote_type == "SingleQuote":
-				return f'\'{tex(e.content, i, 0, "")}\''
-			elif e.quote_type == "DoubleQuote":
-				return f'"{tex(e.content, i, 0, "")}"'
-			else:
-				return f'"{tex(e.content, i, 0, "")}"'
-
-	if isinstance(e, Code):
-		return f'\\verb`{e.text.replace("`", "backtick")}`'
-
-	if isinstance(e, Figure):
-		return f'\\figure{{{tex(e.content, i, indent_level+1, indent_str)}}}{{{tex(e.caption, i, indent_level+1, indent_str)}}}\n\n'
-	
-	# Figure caption
-	if isinstance(e, Caption):
-		if inlinify(e) is not None:
-			return f'\\figcaption{{{tex(e.content, i, 0, "")}}}'
-
-	if isinstance(e, Math):
-		if e.format == "DisplayMath":
-			return f'$${e.text}$$\n'
-		else:
-			return f'${e.text}$'
-	
-	# Footnote
-	if isinstance(e, Note):
-		tag = "fn"
-		if inlinify(e) is not None:
-			return f'\\fn{{{tex(inlinify(e), i, 0, "")}}}'
-
-	if isinstance(e, Table):
-		aligns = {
-			"AlignLeft": "\\quad#\\quad\\hfil",
-			"AlignRight": "\\quad\\hfil#\\quad",
-			"AlignCenter": "\\quad\\hfil#\\hfil\\quad",
-			"AlignDefault": "\\quad#\\quad\\hfil"
-		}
-		text = "\strut"+"&".join([aligns[col[0]] for col in e.colspec])+"\cr\n"
-		text += tex(e.head.content, i, 0, "")
-		text += "\\noalign{\\hrule}\n"
-		text += tex(e.content[0].content, i, 0, "")
-		text += "\\noalign{\\hrule}\n"
-		text += tex(e.foot.content, i, 0, "")
-		return "\\vskip1em\n\\halign{"+text+"}\n\\vskip1em\n"
-		# FIXME: Implement rowspan
-	
-	if isinstance(e, TableRow):
-		return "&".join([("\\multispan"+str(cell.colspan)+" " if cell.colspan > 1 else "")+tex(cell.content, i, 0, "") for cell in e.content])+"\cr\n"
-
-	if isinstance(e, RawInline):
-		if e.format == "tex":
-			return e.text
-		else:
-			return ""
-
-	if isinstance(e, RawBlock):
-		if e.format == "tex":
-			return f'{e.text}\n'
-		else:
-			return ""
-	
-	# See https://pandoc.org/MANUAL.html#line-blocks
-	if isinstance(e, LineBlock):
-		return f'{tex(e.content, i, indent_level+1, indent_str)}\n'
-
-	if isinstance(e, LineItem):
-		return tex(e.content, i, 0, "") + ("\\\\\n" if e.next else "\n")
-
-	if type(e) is Div:
-		return f'{tex(e.content, i, indent_level+1, indent_str)}'
-
-	if isinstance(e, Doc):
-		return tex(e.content, i, indent_level, indent_str)+"\n\\bye" # Is having the \bye a bad idea here?
-
-
-	# Non-overriding elements, they get generated using the template at the end
-	# of this function
-	if isinstance(e, BulletList):
-		tag = "list"
-		open = ""
-		arguments = "{o}"
-		close = "\\endlist"
-
-	elif isinstance(e, OrderedList):
-		tag = "list"
-		open = ""
-		styles = {
-			"DefaultStyle": "n",
-			"Decimal": "n",
-			"LowerRoman": "i",
-			"UpperRoman:": "I",
-			"LowerAlpha": "a",
-			"UpperAlpha": "A"
-		}
-		style = styles[e.style]
-		delimiters = {
-			"DefaultDelim": f"{style}.",
-			"Period": f"{style}.",
-			"OneParen": f"{style})",
-			"TwoParens": f"({style})"
-		}
-		style = delimiters[e.delimiter]
-		arguments = f"{{{style}}}"
-		close = "\\endlist"
-		# FIXME: Starting number of list
-
-	elif isinstance(e, ListItem):
-		tag = ":"
-
-	elif isinstance(e, Link):
-		if len(e.content) == 1 and isinstance(e.content[0], Str) and e.content[0].text == e.url:
-			tag = "url"
-		else:
-			tag = "linkurl"
-			arguments = f'{{{e.url}}}'
-
-	elif isinstance(e, Group):
-		tag = "begingroup"
-		open = ""
-		if "lang" in e.metadata and e.metadata["lang"] is not None:
-			open = "\\language"+e.metadata["lang"]
-		close = "\\endgroup"
-	
-	# The default which all non-overriding elements get generated by. This
-	# includes elements, which were not explicitly mentioned in this function,
-	# e. g. Strong, Emph...
-
-	if isinstance(e, Inline):
-		return f'\\{tag}{arguments}{open}{content_head}{tex(e.content, i, 0, "") if hasattr(e, "_content") else ""}{e.text if hasattr(e, "text") else ""}{content_foot}{close}'
-
-	out_str = ""
-	out_str = f"\\{tag}{arguments}{open}\n"
-	out_str += content_head
-	if hasattr(e, "_content"):
-		out_str += tex(e.content, i, indent_level+1, indent_str)
-	if hasattr(e, "text"):
-		out_str += e.text
-	out_str += f"{content_foot}\n{close}\n\n"
-
-	return out_str

src/formatitko/tex_generator.py

@@ -0,0 +1,366 @@
+from panflute import Element, ListContainer, Inline, Block
+from panflute import Cite, Code, Emph, Image, LineBreak, Link, Math, Note, Quoted, RawInline, SmallCaps, SoftBreak, Space, Span, Str, Strikeout, Strong, Subscript, Superscript, Underline
+from panflute import BlockQuote, BulletList, Citation, CodeBlock, Definition, DefinitionItem, DefinitionList, Div, Figure, Header, HorizontalRule, LineBlock, LineItem, ListItem, MetaBlocks, MetaBool, MetaInlines, MetaList, MetaMap, MetaString, Null, OrderedList, Para, Plain, RawBlock, Table, TableBody, TableFoot, TableHead
+from panflute import TableRow, TableCell, Caption, Doc
+from panflute import MetaValue
+from typing import Union
+
+import os
+
+from .output_generator import OutputGenerator
+from .images import ImageProcessor, ImageProcessorNamespaceSearcher
+
+from .whitespace import NBSP
+from .elements import FQuoted
+from .context import Group, InlineGroup, BlockGroup, Context
+from .util import inlinify
+
+class UCWTexGenerator(OutputGenerator):
+	imageProcessor: ImageProcessor
+	_bold: int
+	_italic: int
+
+	def __init__(self, output_file, imageProcessor: ImageProcessor, *args, **kwargs):
+		self.imageProcessor = imageProcessor
+		self._bold = 0
+		self._italic = 0
+		super().__init__(output_file, *args, **kwargs)
+
+	def escape_special_chars(self, text: str) -> str:
+		text = text.replace("&", r"\&")
+		text = text.replace("%", r"\%")
+		text = text.replace("$", r"\$")
+		text = text.replace("#", r"\#")
+		text = text.replace("_", r"\_")
+		text = text.replace("{", r"\{")
+		text = text.replace("}", r"\}")
+		text = text.replace("~", r"\textasciitilde{}")
+		text = text.replace("^", r"\textasciicircum{}")
+		text = text.replace("\\", r"\textbackslash{}")
+		text = text.replace(" ", "~") # We use unicode no-break spaces to force nbsp in output
+		text = text.replace("​", "")
+		return text
+
+	def generate(self, e: Union[Element, ListContainer]):
+		if hasattr(e, "attributes") and "only" in e.attributes and e.attributes["only"] != "tex":
+			return
+		super().generate(e)
+
+	def writepar(self, text: str):
+		self.ensure_empty(2)
+		self.writeln(text)
+		self.ensure_empty(2)
+
+	def generate_Null(self, e: Null):
+		pass
+
+	def generate_LineBreak(self, e: LineBreak):
+		self.write(r"\\")
+		self.endln()
+
+	def generate_Para(self, e: Para):
+		self.ensure_empty(2)
+		self.generate(e.content)
+		self.ensure_empty(2)
+
+	def generate_HorizontalRule(self, e: HorizontalRule):
+		self.writepar(r"\vskip5pt\hrule\hfil\vskip5pt{}")
+
+	def generate_Doc(self, e: Doc):
+		self.writeln(r"\input ucwmac2.tex")
+		self.writeln(r"\ucwmodule{ofs}")
+		self.writeln(r"\ucwmodule{verb}")
+		self.writeln(r"\ucwmodule{link}")
+		self.writeln(r"\input formatitko.tex")
+		self.generate(e.content)
+		self.writeln(r"\bye")
+
+	def get_language_macro(self, lang: str):
+		if lang == "cs":
+			return r"\chyph\lefthyphenmin=2\righthyphenmin=2{}"
+		elif lang == "sk":
+			return r"\shyph\lefthyphenmin=2\righthyphenmin=2{}"
+		elif lang == "en":
+			return r"\ehyph\lefthyphenmin=2\righthyphenmin=2{}"
+		else:
+			return ""
+
+	def generate_InlineGroup(self, e: InlineGroup):
+		self.write(r"{")
+		self.write(self.get_language_macro(self.context.get_metadata("lang")))
+		self.generate(e.content)
+		self.write(r"}")
+
+	def generate_BlockGroup(self, e: BlockGroup):
+		self.writeln(r"\begingroup")
+		self.indent_more()
+		self.writeln(self.get_language_macro(self.context.get_metadata("lang")))
+		self.generate(e.content)
+		self.indent_less()
+		self.writeln(r"\endgroup")
+	
+	def generate_Header(self, e: Header):
+		self.ensure_empty(2)
+		self.write("\\"+"sub"*(e.level-1)+"section{")
+		self.generate(e.content)
+		self.write(r"}")
+		self.ensure_empty(2)
+
+	def generate_Image(self, e: Image):
+		url = e.url
+
+		additional_args = self.get_image_processor_args(e.attributes)
+		additional_args["context"] = self.context
+
+		# The directory of the current file relative to the current working directory
+		source_dir = self.context.dir
+		# The directory of the current file relative to the md file we were called on
+		rel_dir = self.context.rel_dir
+
+		searcher = self.imageProcessor.get_searcher_by_path(url, rel_dir, source_dir)
+		url = self.imageProcessor.get_path_without_namespace(url)
+
+		_, ext = os.path.splitext(url)
+		ext = ext[1:]
+
+		# Conversions between various formats.
+		if ext in ["pdf", "png", "jpeg"]:
+			# Even supported elements have to be 'converted' because the
+			# processing contains finding and moving them to the cache
+			# directory.
+			url = self.imageProcessor.process_image(url, ext, searcher, **additional_args)
+		elif ext in ["svg"]: # FIXME
+			url = self.imageProcessor.process_image(url, "pdf", searcher, **additional_args)
+		elif ext in ["epdf"]:
+			url = self.imageProcessor.process_image(url, "pdf", searcher, **additional_args)
+		elif ext in ["jpg"]:
+			url = self.imageProcessor.process_image(url, "jpeg", searcher, **additional_args)
+		else:
+			url = self.imageProcessor.process_image(url, "pdf", searcher, **additional_args)
+
+		url = searcher.get_cache_searcher().find_image(url)
+		width = ""
+		if "width" in e.attributes:
+			width = e.attributes["width"]
+			# 50% → 0.5\hsize
+			if e.attributes["width"][-1] == "%":
+				width = str(int(e.attributes["width"][:-1])/100) + "\\hsize"
+			width = "width " + width
+
+		if isinstance(e.parent.parent, Figure):
+			self.writeln(f'\\putimage{{{width}}}{{{url}}}')
+		else:
+			self.writepar(f'\\putimage{{{width}}}{{{url}}}')
+
+	def generate_Code(self, e: Code):
+		self.write(r"\verb`")
+		self.write(e.text)
+		self.write(r"`")
+
+	def generate_Figure(self, e: Figure):
+		self.ensure_empty(2)
+		self.writeln(r"\vskip5pt")
+		self.writeln(r"\centerline{")
+		self.indent_more()
+		self.generate(e.content)
+		self.indent_less()
+		self.writeln(r"}")
+		self.writeln(r"\centerline{")
+		self.indent_more()
+		self.generate(e.caption)
+		self.indent_less()
+		self.writeln(r"}")
+		self.writeln(r"\vskip5pt{}")
+		self.ensure_empty(2)
+
+	def generate_Emph(self, e: Emph):
+		if self._bold > 0:
+			self.write(r"{\bi{}")
+		else:
+			self.write(r"{\I{}")
+		self._italic+=1
+		self.generate(e.content)
+		self._italic-=1
+		self.write(r"}")
+
+	def generate_Strong(self, e: Strong):
+		if self._italic > 0:
+			self.write(r"{\bi{}")
+		else:
+			self.write(r"{\bf{}")
+		self._bold+=1
+		self.generate(e.content)
+		self._bold-=1
+		self.write(r"}")
+
+	def generate_Caption(self, e: Caption):
+		self.generate_Emph(e)
+	
+	def generate_Math(self, e: Math):
+		if e.format == "DisplayMath":
+			self.ensure_empty(2)
+			self.writeraw("$$")
+			self.writeraw(e.text.strip())
+			self.writeraw("$$")
+			self.ensure_empty(2)
+		else:
+			self.write("$")
+			self.write(e.text)
+			self.write("$")
+
+	def generate_Note(self, e: Note):
+		self.write(r"\fn{")
+		self.generate(inlinify(e))
+		self.write(r"}")
+
+	def generate_Table(self, e: Table):
+		aligns = {
+			"AlignLeft": r"\quad#\quad\hfil",
+			"AlignRight": r"\quad\hfil#\quad",
+			"AlignCenter": r"\quad\hfil#\hfil\quad",
+			"AlignDefault": r"\quad#\quad\hfil"
+		}
+		self.writeln(r"\vskip1em")
+		self.writeln(r"\halign{\strut"+"&".join([aligns[col[0]] for col in e.colspec])+r"\cr")
+		self.indent_more()
+		self.generate(e.head.content)
+		self.writeln(r"\noalign{\hrule}")
+		self.generate(e.content[0].content)
+		self.writeln(r"\noalign{\hrule}")
+		self.generate(e.foot.content)
+		self.indent_less()
+		self.writeln("}")
+		self.writeln(r"\vskip1em")
+
+	def generate_TableRow(self, e: TableRow):
+		for cell in e.content:
+			if cell.colspan > 1:
+				self.write(r"\multispan"+str(cell.colspan)+"{} ")
+			self.generate(cell.content)
+			if cell.next:
+				self.write(" & ")
+		self.write(r"\cr")
+		self.endln()
+
+	def generate_RawInline(self, e: RawInline):
+		if e.format == "tex":
+			self.write(e.text)
+
+	def generate_RawBlock(self, e: RawBlock):
+		if e.format == "tex":
+			self.writeraw(e.text)
+
+	def generate_Plain(self, e: Plain):
+		self.generate(e.content)
+
+	def generate_Span(self, e: Span):
+		self.generate(e.content)
+
+	def generate_CodeBlock(self, e: CodeBlock):
+		self.writeln(r"\verbatim{")
+		self.writeraw(e.text)
+		self.writeln(r"}")
+
+	def generate_Div(self, e: Div):
+		self.generate(e.content)
+
+	def generate_LineBlock(self, e: LineBlock):
+		self.writeln()
+		self.generate(e.content)
+		self.writeln()
+	
+	def generate_LineItem(self, e: LineItem):
+		self.generate(e.content)
+		if e.next:
+			self.write(r"\\")
+		self.endln()
+
+	def generate_BulletList(self, e: BulletList):
+		self.ensure_empty(2)
+		self.writeln(r"\list{o}")
+		self.indent_more()
+		self.generate(e.content)
+		self.indent_less()
+		self.write(r"\endlist")
+		self.ensure_empty(2)
+
+	def generate_OrderedList(self, e: OrderedList):
+		self.ensure_empty(2)
+		styles = {
+			"DefaultStyle": "n",
+			"Decimal": "n",
+			"LowerRoman": "i",
+			"UpperRoman:": "I",
+			"LowerAlpha": "a",
+			"UpperAlpha": "A"
+		}
+		style = styles[e.style]
+		delimiters = {
+			"DefaultDelim": f"{style}.",
+			"Period": f"{style}.",
+			"OneParen": f"{style})",
+			"TwoParens": f"({style})"
+		}
+		style = delimiters[e.delimiter]
+		self.writeln(r"\list{"+style+r"}")
+		self.indent_more()
+		self.generate(e.content)
+		self.indent_less()
+		self.writeln(r"\endlist")
+		self.ensure_empty(2)
+	
+	def generate_ListItem(self, e: ListItem):
+		self.endln()
+		self.write(r"\:")
+		self.generate(e.content)
+		self.endln()
+
+	def generate_BlockQuote(self, e: BlockQuote):
+		self.writeln(r"\blockquote{")
+		self.indent_more()
+		self.generate(e.content)
+		self.indent_less()
+		self.writeln(r"}")
+
+	def generate_Link(self, e: Link):
+		if len(e.content) == 1 and isinstance(e.content[0], Str) and e.content[0].text == e.url:
+			self.write(r"\url{")
+		else:
+			self.write(r"\linkurl{"+e.url+r"}{")
+		self.generate(e.content)
+		self.write(r"}") # }
+
+	def generate_Subscript(self, e: Subscript):
+		self.write(r"\subscript{")
+		self.generate(e.content)
+		self.write(r"}")
+
+	def generate_Superscript(self, e: Superscript):
+		self.write(r"\superscript{")
+		self.generate(e.content)
+		self.write(r"}")
+
+	def generate_simple_tag(self, e: Union[Element, None] = None, tag: str = "", attributes: Union[dict[str, str], None] = None, content: Union[ListContainer, Element, list[Union[Element, ListContainer]], str, None] = None, inline: Union[bool, None] = None):
+		print("dumbass: ", type(e))
+
+	# These are also disabled in pandoc so they shouldn't appear in the AST at all.
+	def generate_Citation(self, e: Citation):
+		self.writeln("% FIXME: Citations not implemented")
+
+	def generate_Cite(self, e: Cite):
+		self.writeln("% FIXME: Cites not implemented")
+
+	def generate_Definition(self, e: Definition):
+		self.writeln("% FIXME: Definitions not implemented")
+
+	def generate_DefinitionItem(self, e: DefinitionItem):
+		self.writeln("% FIXME: DefinitionItems not implemented")
+
+	def generate_DefinitionList(self, e: DefinitionList):
+		self.writeln("% FIXME: DefinitionLists not implemented")
+
+	def generate_Underline(self, e: Underline):
+		self.writeln("% FIXME: Underlines not implemented")
+
+	def generate_Strikeout(self, e: Strikeout):
+		self.writeln("% FIXME: Strikeouts not implemented")

src/formatitko/transform.py

@@ -1,176 +0,0 @@
-from panflute import Element, Div, Span, Quoted, Image, CodeBlock, Str, MetaInlines, MetaString, MetaBool, RawBlock
-import re
-import os
-
-# Import local files
-from .whitespace import Whitespace, NBSP, bavlna
-from .util import nullify, import_md
-from .context import Context, BlockGroup
-from .command import Command, BlockCommand, InlineCommand
-from .command_util import handle_command_define, parse_command
-from .elements import FQuoted
-
-
-import warnings
-warnings.warn("The transform function has been deprecated, is left only for reference and will be removed in future commits. TransformProcessor should be used in its place.", DeprecationWarning)
-
-# This is where tha magic happens. This function transforms a single element,
-# to transform the entire tree, panflute's walk should be used.
-def transform(e: Element, c: Context) -> Element:
-
-	warnings.warn("The transform function has been deprecated, is left only for reference and will be removed in future commits. TransformProcessor should be used in its place.", DeprecationWarning)
-	# Determine if this space should be non-breakable. See whitespace.py.
-	if isinstance(e, Whitespace) and bavlna(e, c):
-		e = NBSP()
-
-	if hasattr(e, "attributes"):
-		# `if` attribute. Only show this element if flag is set.
-		if "if" in e.attributes:
-			if not c.is_flag_set(e.attributes["if"]):
-				return nullify(e)
-		# `ifn` attribute. Only show this element if flag is NOT set
-		if "ifn" in e.attributes:
-			if c.is_flag_set(e.attributes["ifn"]):
-				return nullify(e)
-
-		# There are multiple ways to call a command so we turn it into a
-		# unified element first and then call it at the end. This handles the
-		# []{c=commandname} and
-		# :::{c=commandname}
-		# :::
-		# syntax.
-		if (isinstance(e, Div) or isinstance(e, Span)) and "c" in e.attributes:
-			if isinstance(e, Div):
-				e = BlockCommand(*e.content, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
-			else:
-				e = InlineCommand(*e.content, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
-
-		# Isolated subdocuments using Group and a different Context.  Can be
-		# separate files (using attribute `partial`) or be inline using the
-		# following syntax:
-		# ```markdown {.group}
-		# * file content *
-		# ```
-		# Both can contain their own metadata in a FrontMatter (YAML header)
-		if (isinstance(e, Div) and "partial" in e.attributes)\
-		  or (isinstance(e, CodeBlock) and "markdown" in e.classes and "group" in e.classes):
-			if isinstance(e, Div):
-				if not c.trusted: # If we're in an untrusted context, we shouldn't allow inclusion of files outside the PWD.
-					full_path = os.path.abspath(c.dir + "/" + e.attributes["partial"])
-					pwd = os.path.abspath(".")
-					if os.path.commonpath([full_path, pwd]) != os.path.commonpath([pwd]):
-						return nullify(e)
-				text = open(c.dir + "/" + e.attributes["partial"], "r").read()
-				path = c.dir + "/" + e.attributes["partial"]
-			else:
-				text = e.text
-				path = c.path
-			if "type" in e.attributes and e.attributes["type"] in ["tex", "html"]:
-				e = RawBlock(text, e.attributes["type"])
-			else:
-				includedDoc = import_md(text)
-				trusted = True
-				if "untrusted" in e.attributes and (e.attributes["untrusted"] == True or e.attributes["untrusted"] == 'True'):
-					trusted = False
-				if not c.trusted:
-					trusted = False
-				nContext = Context(includedDoc, path, c, trusted=trusted)
-				language = includedDoc.get_metadata("lang")
-				includedDoc = includedDoc.walk(transform, nContext)
-				e = BlockGroup(*includedDoc.content, context=nContext, metadata={"lang": language})
-
-	# Transform panflute's Quoted to custom FQuoted, see above.
-	if isinstance(e, Quoted):
-		quote_styles = {
-			"cs": "cs",
-			"en": "en",
-			"sk": "cs",
-			None: None
-		}
-		e = FQuoted(*e.content, quote_type=e.quote_type, style=quote_styles[c.get_metadata("lang")])
-	
-	if isinstance(e, Image):
-		# Pass down the directory of the current source file for finding image
-		# files.
-		e.attributes["source_dir"] = c.dir
-		# Pass down "no-srcset" metadatum as attribute down to images.
-		if not "no-srcset" in e.attributes:
-			e.attributes["no-srcset"] = c.get_metadata("no-srcset") if c.get_metadata("no-srcset") is not None else False
-
-	# Pass down metadata 'highlight' and 'highlight_style' as attribute to CodeBlocks
-	if isinstance(e, CodeBlock):
-		if not "highlight" in e.attributes:
-			e.attributes["highlight"] = c.get_metadata("highlight") if c.get_metadata("highlight") is not None else True
-		if not "style" in e.attributes:
-			e.attributes["style"] = c.get_metadata("highlight-style") if c.get_metadata("highlight-style") is not None else "default"
-			e.attributes["noclasses"] = False
-			# I think this is supposed to enable inline styles for highlighting when the style differs from the document, but it clearly doesn't work. a) HTML_generator never accesses it and b) Only the top-level document contains a style so you have to ask the top level context, not the current context.
-		else:
-			e.attributes["noclasses"] = True
-
-	# Execute python code inside source code block. Works the same as commands.
-	# Syntax:
-	# ```python {.run}
-	# print("woo")
-	# ```
-	if isinstance(e, CodeBlock) and hasattr(e, "classes") and "python" in e.classes and "run" in e.classes:
-		if not c.trusted:
-			return nullify(e)
-		command_output = parse_command(e.text)(BlockCommand(), c)
-		e = BlockCommand().replaceSelf(*([] if command_output is None else command_output))
-		e = e.walk(transform, c)
-
-	# Command defines for calling using BlockCommand and InlineCommand. If
-	# redefine is used instead of define, the program doesn't check if the
-	# command already exists.
-	# Syntax:
-	# ```python {define=commandname}
-	# print(wooo)
-	# ```
-	if isinstance(e, CodeBlock) and hasattr(e, "classes") and "python" in e.classes and hasattr(e, "attributes")\
-		and ("define" in e.attributes or "redefine" in e.attributes):
-		if not c.trusted:
-			return nullify(e)
-		e = handle_command_define(e, c)
-
-	## Shorthands
-	# Shorter (and sometimes the only) forms of certain features
-	if isinstance(e, Span) and len(e.content) == 1 and isinstance(e.content[0], Str):
-		## Handle special command shorthand [!commandname]{}
-		if re.match(r"^![\w.]+$", e.content[0].text):
-			e = InlineCommand(identifier=e.identifier, classes=e.classes, attributes={**e.attributes, "c": e.content[0].text[1:]})
-
-		## Handle import [#path/file.md]{}
-		# This is the exact opposite of partials. We take the commands, flags
-		# and metadata but drop the content.
-		elif re.match(r"^#.+$", e.content[0].text):
-			importedDoc = import_md(open(c.dir + "/" + e.content[0].text[1:], "r").read())
-			importedDoc.walk(transform, c)
-			return nullify(e)
-
-		## Handle metadata print [$key1.key2]{}
-		# This is a shorthand for just printing the content of some metadata.
-		elif re.match(r"^\$[\w.]+$", e.content[0].text):
-			val = c.get_metadata(e.content[0].text[1:], False)
-			if isinstance(val, MetaInlines):
-				e = Span(*val.content)
-				e = e.walk(transform, c)
-			elif isinstance(val, MetaString):
-				e = Span(Str(val.string))
-			elif isinstance(val, MetaBool):
-				e = Span(Str(str(val.boolean)))
-			else:
-				raise TypeError(f"Cannot print value of metadatum '{e.content[0].text[1:]}' of type '{type(val)}'")
-
-	## Execute commands
-	# panflute's walk function transforms the children first, then the root
-	# element, so the content the command receives is already transformed.
-	# The output from the command is then transformed manually again.
-	if isinstance(e, Command):
-		if not c.get_command(e.attributes["c"]):
-			raise NameError(f"Command not defined '{e.attributes['c']}'.")
-		command_output = c.get_command(e.attributes["c"])(e, c)
-		e = e.replaceSelf(*command_output)
-		e = e.walk(transform, c)
-
-	return e

src/formatitko/transform_processor.py

@@ -3,6 +3,7 @@ from panflute import Cite, Code, Emph, Image, LineBreak, Link, Math, Note, Quote
 from panflute import BlockQuote, BulletList, Citation, CodeBlock, Definition, DefinitionItem, DefinitionList, Div, Figure, Header, HorizontalRule, LineBlock, LineItem, ListItem, MetaBlocks, MetaBool, MetaInlines, MetaList, MetaMap, MetaString, Null, OrderedList, Para, Plain, RawBlock, Table, TableBody, TableFoot, TableHead
 from panflute import TableRow, TableCell, Caption, Doc
 from panflute import MetaValue
+from panflute.containers import attach
 from typing import Union, Callable
 from types import ModuleType
 
@@ -18,17 +19,12 @@ from .context import Group, InlineGroup, BlockGroup
 from .util import nullify, import_md
 from .context import Context, CommandCallable
 from .whitespace import Whitespace, bavlna
-from .command import BlockCommand, InlineCommand, CodeCommand, Command
+from .command import BlockCommand, InlineCommand, CodeCommand, Command, InlineError
 from .command_util import handle_command_define, parse_command
-from .nop_processor import NOPProcessor, ELCl
-
-class DoubleDocError(Exception):
-	"TransformProcessor should only ever see a single Doc."
-	pass
+from .nop_processor import NOPProcessor, ELCl, DoubleDocError
 
 class TransformProcessor(NOPProcessor):
 	
-	context: Union[Context, None] = None
 	root_file_path: str
 	root_highlight_style: str = "default"
 	_command_modules: list[tuple[Union[dict[str, CommandCallable], ModuleType], str]] = []
@@ -44,6 +40,15 @@ class TransformProcessor(NOPProcessor):
 	def add_command_module(self, module: Union[dict[str, CommandCallable], ModuleType], module_name: str=""):
 		self._command_modules.append((module, module_name))
 
+	def init_context(self, e: Doc) -> Context:
+		if self.context is not None:
+			raise DoubleDocError()
+		self.context = Context(e, self.root_file_path)
+		for module, module_name in self._command_modules:
+			self.context.add_commands_from_module(module, module_name)
+		e.content = [BlockGroup(*e.content, context=self.context)]
+		return self.context
+
 	def get_pretransformers(self) -> list[Callable[[ELCl],ELCl]]:
 		return super().get_pretransformers()+[self.handle_if_attribute, self.handle_ifnot_attribute]
 
@@ -61,15 +66,22 @@ class TransformProcessor(NOPProcessor):
 				return nullify(e)
 		return e
 
+	def transform_ListContainer(self, e: ListContainer) -> ListContainer:
+		try:
+			return super().transform_ListContainer(e)
+		except TypeError as err:
+			names = []
+			for el in e:
+				if hasattr(el, "attributes") and "c" in el.attributes:
+					names.append(el.attributes["c"])
+			if len(names) > 0:
+				raise InlineError(f"The command{'s' if len(names) > 1 else ''} {names[0] if len(names) == 1 else names} was called in an Inline way but returned Block content. Put it in a paragraph alone or execute it as a Div using: \n::: {{c={names[0] if len(names) == 1 else '<command_name>'}}}\n:::")
+			else:
+				raise err
 
 	def transform_Doc(self, e: Doc) -> Doc:
-		if self.context is not None:
-			raise DoubleDocError()
-		self.context = Context(e, self.root_file_path)
-		for module, module_name in self._command_modules:
-			self.context.add_commands_from_module(module, module_name)
+		self.init_context(e)
 		e.content = self.transform(e.content)
-		e.content = [BlockGroup(*e.content, context=self.context)]
 		return e
 
 
@@ -87,22 +99,38 @@ class TransformProcessor(NOPProcessor):
 		e.content = self.transform(e.content)
 		# OG now has Context so this is not needed per se, but I'm keeping this here for the handling of attribute > context > default value
 		# Pass down "no-srcset" metadatum as attribute down to images.
-		if not "no-srcset" in e.attributes:
+		if "no-srcset" not in e.attributes:
 			e.attributes["no-srcset"] = self.context.get_metadata("no-srcset") if self.context.get_metadata("no-srcset") is not None else False
+		if "no-img-link" not in e.attributes:
+			e.attributes["no-img-link"] = self.context.get_metadata("no-img-link") if self.context.get_metadata("no-img-link") is not None else False
 		return e
 
-	def create_Group(self, *content, new_context: Context, inline: bool=False) -> Group:
+	def create_Group(self, *content, new_context: Context, replaced:Element, inline: bool=False) -> Group:
 		old_context = self.context
 		self.context = new_context
-		content = self.transform([*content])
-		self.context = old_context
 		if inline:
-			return InlineGroup(*content, context=new_context)
+			g = InlineGroup(*content, context=new_context)
 		else:
-			return BlockGroup(*content, context=new_context)
+			g = BlockGroup(*content, context=new_context)
+		attach(g, replaced.parent, replaced.location, replaced.index)
+		g = self.transform(g)
+		self.context = old_context
+		return g
+
+	def transform_Para(self, e: Para) -> Union[Para, Div]:
+		if len(e.content) == 1 and isinstance(e.content[0], Span):
+			# If the span turns out to be a command, it might return a Div. We should then replace ourselves with the Div
+			span = e.content[0]
+			span = self.transform(span)
+			if isinstance(span, Div):
+				return span
+			else:
+				e.content[0] = span
+				return super().transform_Para(e)
+		else:
+			return super().transform_Para(e)
 
 	def transform_Div(self, e: Div) -> Union[Div, Group, Null, RawBlock]:
-		e.content = self.transform(e.content)
 
 		if "group" in e.classes:
 			# `.group` class for Divs
@@ -110,14 +138,15 @@ class TransformProcessor(NOPProcessor):
 			new_context = Context(Doc(), self.context.path, self.context, trusted=self.context.trusted)
 			for attribute, value in e.attributes.items():
 				new_context.set_metadata(attribute, value)
-			return self.create_Group(*e.content, new_context=new_context)
+			return self.create_Group(*e.content, replaced=e, new_context=new_context)
 
 		if "c" in e.attributes:
 			# Commands can be called multiple ways, this handles the following syntax:
 			# :::{c=commandname}
 			# :::
-			e = BlockCommand(*e.content, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
-			return self.transform(e)
+			command = BlockCommand(*e.content, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
+			attach(command, e.parent, e.location, e.index)
+			return self.transform(command)
 
 		if "partial" in e.attributes:
 			# `partial` attribute
@@ -129,7 +158,9 @@ class TransformProcessor(NOPProcessor):
 				pwd = os.path.abspath(".")
 				if os.path.commonpath([full_path, pwd]) != os.path.commonpath([pwd]):
 					return nullify(e)
-			text = open(self.context.dir + "/" + e.attributes["partial"], "r").read()
+			filename = self.context.dir + "/" + e.attributes["partial"]
+			self.context.add_dep(filename)
+			text = open(filename, "r").read()
 			path = self.context.dir + "/" + e.attributes["partial"]
 			if e.attributes["type"] == "md":
 				includedDoc = import_md(text)
@@ -138,7 +169,7 @@ class TransformProcessor(NOPProcessor):
 					trusted = False
 				if not self.context.trusted:
 					trusted = False
-				return self.create_Group(*includedDoc.content, new_context=Context(includedDoc, path, self.context, trusted=trusted))
+				return self.create_Group(*includedDoc.content, replaced=e, new_context=Context(includedDoc, path, self.context, trusted=trusted))
 			elif e.attributes["type"] in ["tex", "html"]:
 				return RawBlock(text, e.attributes["type"])
 
@@ -159,10 +190,9 @@ class TransformProcessor(NOPProcessor):
 		if "lang" in e.attributes:
 			warnings.warn("To set language in a way formátítko will understand, this Div has to have the `.group` class and be a Group.", UserWarning)
 
-		return e
+		return super().transform_Div(e)
 
 	def transform_Span(self, e: Span) -> Span:
-		e.content = self.transform(e.content)
 
 		if "group" in e.classes:
 			# `.group` class for Spans
@@ -170,19 +200,21 @@ class TransformProcessor(NOPProcessor):
 			new_context = Context(Doc(), self.context.path, self.context, trusted=self.context.trusted)
 			for attribute, value in e.attributes.items():
 				new_context.set_metadata(attribute, value)
-			return self.create_Group(*e.content, new_context=new_context, inline=True)
+			return self.create_Group(*e.content, replaced=e, new_context=new_context, inline=True)
 
 		if "c" in e.attributes:
 			# Commands can be called multiple ways, this handles the following syntax:
 			# []{c=commandname} and
-			e = InlineCommand(*e.content, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
-			return self.transform(e)
+			command = InlineCommand(*e.content, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
+			attach(command, e.parent, e.location, e.index)
+			return self.transform(command)
 
 		if len(e.content) == 1 and isinstance(e.content[0], Str):
 			## Handle special command shorthand [!commandname]{}
 			if re.match(r"^![\w.]+$", e.content[0].text):
-				e = InlineCommand(identifier=e.identifier, classes=e.classes, attributes={**e.attributes, "c": e.content[0].text[1:]})
-				return self.transform(e)
+				command = InlineCommand(identifier=e.identifier, classes=e.classes, attributes={**e.attributes, "c": e.content[0].text[1:]})
+				attach(command, e.parent, e.location, e.index)
+				return self.transform(command)
 
 			## Handle import [#ksp_formatitko as ksp]{}, [#ksp_formatitko]{type=module} or [#path/file.md]{type=md}
 			# Import a python module as commands (type=module, the default) or
@@ -191,7 +223,9 @@ class TransformProcessor(NOPProcessor):
 				if not "type" in e.attributes:
 					e.attributes["type"] = "module"
 				if e.attributes["type"] == "md":
-					importedDoc = import_md(open(self.context.dir + "/" + e.content[0].text[1:], "r").read())
+					filename = self.context.dir + "/" + e.content[0].text[1:]
+					self.context.add_dep(filename)
+					importedDoc = import_md(open(filename, "r").read())
 					self.transform(importedDoc.content)
 				elif e.attributes["type"] == "module":
 					matches = re.match(r"^(\w+)(?: as (\w+))?$", e.content[0].text[1:])
@@ -201,7 +235,9 @@ class TransformProcessor(NOPProcessor):
 					module_name = matches.group(1) if matches.group(2) is None else matches.group(2)
 					self.context.add_commands_from_module(module, module_name)
 				elif e.attributes["type"] == "metadata":
-					data = json.load(open(self.context.dir + "/" + e.content[0].text[1:], "r"))
+					filename = self.context.dir + "/" + e.content[0].text[1:]
+					self.context.add_dep(filename)
+					data = json.load(open(filename, "r"))
 					key = "" if not "key" in e.attributes else e.attributes["key"]
 					self.context.import_metadata(data, key)
 				else:
@@ -213,7 +249,7 @@ class TransformProcessor(NOPProcessor):
 			# This is a shorthand for just printing the content of some metadata.
 			elif re.match(r"^\$[\w.]+$", e.content[0].text):
 				val = self.context.get_metadata(e.content[0].text[1:], False)
-				if isinstance(val, MetaInlines):
+				if isinstance(val, MetaInlines): # TODO: Trust transform for this
 					e = Span(*val.content)
 					e = self.transform(e)
 				elif isinstance(val, MetaString):
@@ -224,19 +260,20 @@ class TransformProcessor(NOPProcessor):
 					raise TypeError(f"Cannot print value of metadatum '{e.content[0].text[1:]}' of type '{type(val)}'")
 				return e
 
-		return e
+		return super().transform_Span(e)
 
 	def transform_CodeBlock(self, e: CodeBlock) -> Union[CodeBlock, Div, Null]:
 		if "markdown" in e.classes and "group" in e.classes:
 			includedDoc = import_md(e.text)
-			return self.create_Group(*includedDoc.content, new_context=Context(includedDoc, self.context.path, self.context, self.context.trusted))
+			return self.create_Group(*includedDoc.content, replaced=e, new_context=Context(includedDoc, self.context.path, self.context, self.context.trusted))
 
 		if "python" in e.classes and "run" in e.classes:
 			if not self.context.trusted:
 				return nullify(e)
-			command_output = parse_command(e.text)(BlockCommand(), self.context)
-			e = BlockCommand().replaceSelf(*([] if command_output is None else command_output))
-			return self.transform(e)
+			command_output = parse_command(e.text)(BlockCommand(), self.context, self)
+			command = BlockCommand().replaceSelf(*([] if command_output is None else command_output))
+			attach(command, e.parent, e.location, e.index)
+			return self.transform(command)
 
 		if "python" in e.classes and ("define" in e.attributes or "redefine" in e.attributes):
 			if not self.context.trusted:
@@ -244,7 +281,9 @@ class TransformProcessor(NOPProcessor):
 			return handle_command_define(e, self.context)
 
 		if "c" in e.attributes:
-			return self.transform(CodeCommand(e.text, identifier=e.identifier, classes=e.classes, attributes=e.attributes))
+			command = CodeCommand(e.text, identifier=e.identifier, classes=e.classes, attributes=e.attributes)
+			attach(command, e.parent, e.location, e.index)
+			return self.transform(command)
 
 		# Pass down metadata 'highlight' and 'highlight_style' as attribute to CodeBlocks
 		# OG now has Context so this is not needed per se, but I'm keeping this here for the handling of attribute > context > default value
@@ -257,9 +296,9 @@ class TransformProcessor(NOPProcessor):
 	def transform_Command(self, e: Command) -> Union[Div, Span]:
 		if not self.context.get_command(e.attributes["c"]):
 			raise NameError(f"Command not defined '{e.attributes['c']}'.")
-		command_output = self.context.get_command(e.attributes["c"])(e, self.context)
+		command_output = self.context.get_command(e.attributes["c"])(e, self.context, self)
 		e = e.replaceSelf(*([] if command_output is None else command_output))
-		return self.transform(e)
+		return e
 
 	def transform_Whitespace(self, e: Whitespace) -> Whitespace:
 		if bavlna(e, self.context):

src/formatitko/util.py

@@ -37,7 +37,7 @@ def parse_string(s: str) -> list[Union[Str, Space]]:
 # we ever want to disable or enable some of panflute's markdown extensions,
 # this is the place to do it.
 def import_md(s: str, standalone: bool=True) -> Union[Doc, list[Element]]:
-	return convert_text(s, standalone=standalone, input_format="markdown-definition_lists-latex_macros")
+	return convert_text(s, standalone=standalone, input_format="markdown-definition_lists-latex_macros", extra_args=["--strip-comments"])
 
 def import_md_list(s: str) -> list[Element]:
 	return import_md(s, standalone=False)

src/formatitko/whitespace.py

@@ -36,9 +36,9 @@ def bavlna(e: Whitespace, c: Context) -> bool:
 		if prevC in operators and nextC in numbers:
 			return True
 
-	if isinstance(e.prev, Math) or isinstance(e.next, Math):
-		# Add no-break spaces around TeX math.
-		return True
+#	if isinstance(e.prev, Math) or isinstance(e.next, Math):
+#		# Add no-break spaces around TeX math.
+#		return True
 
 
 

test/test-files/test-partial.md

@@ -56,7 +56,16 @@ $$
 $$
 
 
-![This is a figure, go figure...](logo.svg){width=25%}\
+<!--There is an inline *emphasis with $math \error$*.-->
+
+<!--
+```python {.run}
+print("bruh")
+raise Exception("Jsem piča")
+```
+-->
+
+![This is a figure, go figure...](logo.svg){width=25%}What
 
 ![This is a figure, go figure...](logo.pdf){width=50%}
 

test/test-top.html

@@ -1,8 +0,0 @@
-<!DOCTYPE html>
-<html>
-<head>
-	<meta name="viewport" content="width=device-width, initial-scale=1">
-	<meta charset='utf-8'>
-	<link rel='stylesheet' href='https://cdn.jsdelivr.net/npm/katex@0.16.4/dist/katex.min.css' integrity='sha384-vKruj+a13U8yHIkAyGgK1J3ArTLzrFGBbBc0tDp4ad/EyewESeXE/Iv67Aj8gKZ0' crossorigin='anonymous'>
-</head>
-<body>

test/test.md

@@ -3,12 +3,13 @@ title: 'Wooooo a title'
 subtitle: 'A subtitle'
 are_we_there_yet: False
 lang: "en"
+header-includes: |
+	<style>
+		body {
+			color: forestgreen;
+		}
+	</style>
 ---
-:::: {.header_content}
-::: {partial="test-top.html" type="html"}
-:::
-::::
-
 [#test-files/test-import.md]{type=md}
 
 [#test.json]{type=metadata key=orgs}
@@ -56,6 +57,11 @@ def bruh(no):
     wat
 ```
 
+```python {style=dracula}
+def bruhec(bruzek):
+	hah
+```
+
 Inline `code`
 
 ::::{if=cat}
@@ -193,11 +199,29 @@ ii. wym bro
 ```python {define=bash}
 import subprocess
 c = subprocess.run(["bash", "-c", element.text], stdout=subprocess.PIPE, check=True, encoding="utf-8")
-return [pf.Para(pf.Str(c.stdout))]
+return [pf.CodeBlock(c.stdout)]
 ```
 
 ```bash {c=bash}
-cat /etc/hostname
+cat /etc/os-release
 ```
 
+::: {.group lang=cs}
+```python {.run}
+return processor.transform([
+	*parse_string("V "),
+	pf.Link(pf.Str("odevzdávátku"), url="https://ksp.mff.cuni.cz/z/odevzdavatko/"),
+	*parse_string(" si necháte vygenerovat vstupy a odevzdáte příslušné výstupy. Záleží jen na vás, jak výstupy vyrobíte.")
 
+])
+```
+:::
+
+```html
+<div>
+hahahahaah
+</div>
+
+```
+
+`<div>`