Previously we got a crash, because we were trying to print
a native cmark STRIKETHROUGH node, and the commonmark writer
in cmark-github doesn't support this. Work around this by
using a raw node to add the strikethrough delimiters.
Closes#4038.
* Move as much as possible to the CSS in the template.
* Ensure that all the HTML-based templates (including epub)
contain the CSS for columns.
* Columns default to 50% width unless they are given a width
attribute.
Closes#4028.
The line identifiers are built using the code block's identifier
as a prefix. If the code block has null identifier, we use
"cb1", "cb2", etc.
Closes#4031.
* Remove "width" attribute which is not allowed on div.
* Remove space between `<div class="column">` elements,
since this prevents columns whose widths sum to 100%
(the space takes up space).
Closes#4028.
<annotation> is not allowed inside <body> according to FictionBook2 XML schema. Besides that, the same information is already placed inside <description>.
Related bug: #2424
and other non-HTML formats (`Text.Pandoc.Readers.HTML.htmlTag`).
The parser stopped at the first `>` character, even if it wasn't
the end of the comment.
Closes#4019.
Previously bullet lists interacted in odd way with ordered lists.
For example, bullet lists nested in ordered list had incorrect
indentation. Besides that, indentation with spaces is not rendered
by FBReader and fbless. To avoid this problem, bullet lists are
indented by appending bullets to marker just the same way it is
done for ordered lists.
- Ensure that epub2 is recognized as a non-text format,
so that a template is used.
- Don't include "prefix" attribute for ibooks for epub2.
It doesn't validate.
- Fix stylesheet paths; previously we had an incorrect
stylesheet path for the cover page and nav page.
According to FB2 XML schema <empty-line /> cannot be placed inside
<p>. Linux FBReader can't display such paragraphs, e.g. any "loose"
lists produced by pandoc prior to this commit. Besides that,
FB2 writer placed <p> inside <p> when writing nested lists,
this commit fixes the bug.
Also this commit removes leading non-breaking space from ordered
lists for consistency with bullet lists.
Definition lists are not affected at all.
* Added underlineSpan builder function. This can be easily updated if needed. The purpose is for Readers to transform underlines consistently.
* Docx Reader: Use underlineSpan and update test
* Org Reader: Use underlineSpan and add test
* Textile Reader: Use underlineSpan and add test case
* Txt2Tags Reader: Use underlineSpan and update test
* HTML Reader: Use underlineSpan and add test case
if the source name is `source` (i.e., the top level). Otherwise
results will be misleading, since what readM does is to look
up the source position of the error in the *original* input,
which may not match the input used by parseFromString or
in parsing from an included file.
Closes#3865. Not a great fix, maybe there's something better
that could be done, but this should at least avoid misleading
messages.
This should help with #3979.
In my test, I got a data uri for data-background. But it didn't
actually work in the browser (the background image didn't show).
Not sure whether this is a problem in reveal.js or a problem
in pandoc...
Previously `\include` wouldn't work if the included file
contained, e.g., a begin without a matching end.
We've changed the Tok type so that it stores a full SourcePos,
rather than just a line and column. So tokens keeep track
of the file they came from. This allows us to use a simpler
method for includes, which doesn't require parsing the included
document as a whole.
Closes#3971.
Now all the guts of openURL have been put into openURL from
Class. openURL is now sensitive to stRequestHeaders in CommonState
and will add these custom headers when making a request.
It no longer looks at the USER_AGENT environment variable,
since you can now set the `User-Agent` header directly.
* In Options.HTMLMathMethod, the KaTeX contsructor now takes only
one string (for the KaTeX base URL), rather than two [API change].
* The default URL has been updated to the latest version.
* The autoload script is now loaded by default.
The pipe command is wrapped in a lua function, throwing a lua error if
the command returns with an error. A wrapper is needed as Haskell
functions exposed to lua may not throw lua errors due to limitations of
hslua.
The error handling is written such that a table can be returned as an
error object in the future. This is potentially useful when finer
control is required while catching the error in lua code. Current
limitations of hslua require error objects to be strings.
Now 'fetch' simply fetches content and mime type.
A new 'hashname' function is provided to get a filename based
on the sha1 hash of the contents and the mime type.
Removed `writerSourceURL` from `WriterOptions` (API change).
Added `stSourceURL` to `CommonState`.
It is set automatically by `setInputFiles`.
Text.Pandoc.Class now exports `setInputFiles`, `setOutputFile`.
The type of `getInputFiles` has changed; it now returns `[FilePath]`
instead of `Maybe [FilePath]`.
Functions in Class that formerly took the source URL as a parameter
now have one fewer parameter (`fetchItem`, `downloadOrRead`,
`setMediaResource`, `fillMediaBag`).
Removed `WriterOptions` parameter from `makeSelfContained` in
`SelfContained`.
This is standard for lua scripts, and I see no reason
to depart from the standard here.
Also, "arg" is now pushed onto the stack before the script
is loaded. Previously it was not, and thus "PandocParameters"
was not available at the top level.
The org reader was updated to match current org-mode behavior: the set
of characters which are acceptable to occur as the first or last
character in an org emphasis have been changed and now allows all
non-whitespace chars at the inner border of emphasized text (see
`org-emphasis-regexp-components`).
Fixes: #3933
* Options: Added readerStripComments to ReaderOptions.
* Added `--strip-comments` command-line option.
* Made `htmlTag` from the HTML reader sensitive to this feature.
This affects Markdown and Textile input.
Closes#2552.
* Rename --latex-engine to --pdf-engine
* In `Text.Pandoc.Options.WriterOptions`, rename `writerLaTeXEngine` to `writerPdfEngine` and `writerLaTeXArgs` to `writerPdfArgs`.
* Add support for `weasyprint` and `prince`, in addition to `wkhtmltopdf`, for PDF generation via HTML (closes#3906).
* `Text.Pandoc.PDF.html2pdf`: use stdin instead of intermediate HTML file
We now use the default.latex template for both latex and beamer.
It contains conditionals for the beamer-specific things.
`pandoc -D beamer` will return this template.
Div's are difficult to translate into org syntax, as there are multiple
div-like structures (drawers, special blocks, greater blocks) which all
have their advantages and disadvantages. Previously pandoc would
use raw HTML to preserve the full div information; this was rarely
useful and resulted in visual clutter. Div-rendering was changed to
discard the div's classes and key-value pairs if there is no natural way
to translate the div into an org structure.
Closes: #3771
Previously pandoc would sometimes combine two line blocks separated by blanks, and ignore trailing blank lines within the line block.
Test is checked to be consisted with http://rst.ninjs.org/
This change makes it possible to define a catch-all function using lua's
metatable lookup functionality.
function catch_all(el)
…
end
return {
setmetatable({}, {__index = function(_) return catch_all end})
}
A further effect of this change is that the map with filter functions
now only contains functions corresponding to AST element constructors.
Closes#3511.
Previously pandoc used the four-space rule: continuation paragraphs,
sublists, and other block level content had to be indented 4
spaces. Now the indentation required is determined by the
first line of the list item: to be included in the list item,
blocks must be indented to the level of the first non-space
content after the list marker. Exception: if are 5 or more spaces
after the list marker, then the content is interpreted as an
indented code block, and continuation paragraphs must be indented
two spaces beyond the end of the list marker. See the CommonMark
spec for more details and examples.
Documents that adhere to the four-space rule should, in most cases,
be parsed the same way by the new rules. Here are some examples
of texts that will be parsed differently:
- a
- b
will be parsed as a list item with a sublist; under the four-space
rule, it would be a list with two items.
- a
code
Here we have an indented code block under the list item, even though it
is only indented six spaces from the margin, because it is four spaces
past the point where a continuation paragraph could begin. With the
four-space rule, this would be a regular paragraph rather than a code
block.
- a
code
Here the code block will start with two spaces, whereas under
the four-space rule, it would start with `code`. With the four-space
rule, indented code under a list item always must be indented eight
spaces from the margin, while the new rules require only that it
be indented four spaces from the beginning of the first non-space
text after the list marker (here, `a`).
This change was motivated by a slew of bug reports from people
who expected lists to work differently (#3125, #2367, #2575, #2210,
#1990, #1137, #744, #172, #137, #128) and by the growing prevalance
of CommonMark (now used by GitHub, for example).
Users who want to use the old rules can select the `four_space_rule`
extension.
* Added `four_space_rule` extension.
* Added `Ext_four_space_rule` to `Extensions`.
* `Parsing` now exports `gobbleAtMostSpaces`, and the type
of `gobbleSpaces` has been changed so that a `ReaderOptions`
parameter is not needed.
Previously only `[-@roe]` (with brackets) was recognized as
suppress-author, and `-@roe` was treated the same as `@roe`.
Closesjgm/pandoc-citeproc#237.
We now allow default output to stdout when it can be
determined that the output is being piped. (On Windows,
as mentioned before, this can't be determined.)
Using '-o -' forces output to stdout regardless.
Acronyms are not resolved by the reader, but acronym and glossary information is put into attributes on Spans so that they can be processed in filters.
Previously, for binary formats, output to stdout was disabled
unless we could detect that the output was being piped (and not
sent to the terminal). Unfortunately, such detection is not
possible on Windows, leaving windows users no way to pipe binary
output. So we have changed the behavior in the following way:
* If the -o option is not used, binary output is never sent
to stdout by default; instead, an error is raised.
* IF '-o -' is used, binary output is sent to stdout, regardless
of whether it is being piped. This works on Windows too.
Instead, just try running it and raise the exception if it
isn't found at that point.
This improves things for users of Cygwin on Windows, where
the executable won't be found by `findExecutable` unless
`.exe` is added.
The same exception is raised as before, but at a later
point.
Closes#3819.
The structure expected is:
<div class="columns">
<div class="column" width="40%">
contents...
</div>
<div class="column" width="60%">
contents...
</div>
</div>
Support has been added for beamer and all HTML slide formats.
Closes#1710.
Note: later we could add a more elegant way to create
this structure in Markdown than to use raw HTML div elements.
This would come for free with a "native div syntax" (#168).
Or we could devise something specific to slides
@ is commonly used in macros using `\makeatletter`.
Ideally we'd make the tokenizer sensitive to `\makeatletter`
and `\makeatother`, but until then this seems a good change.
Stack instances for common data types are now provides by hslua. The
instance for Either was useful only for a very specific case; the
function that was using the `ToLuaStack Either` instance was rewritten
to work without it.
Closes: #3805
Raw table accessing functions never call back into haskell, which allows
the compiler to use more aggressive optimizations. This improves lua
filter performance considerably (⪆5% speedup).
We assume that comments are defined as parsed by the
docx reader:
I want <span class="comment-start" id="0" author="Jesse Rosenthal"
date="2016-05-09T16:13:00Z">I left a comment.</span>some text to
have a comment <span class="comment-end" id="0"></span>on it.
We assume also that the id attributes are unique and properly
matched between comment-start and comment-end.
Closes#2994.
Previously they would be transmitted to the template without
any escaping.
Note that `--M title='*foo*'` yields a different result from
---
title: *foo*
---
In the latter case, we have emphasis; in the former case, just
a string with literal asterisks (which will be escaped
in formats, like Markdown, that require it).
Closes#3792.
* readDataFile, readDefaultDataFile, getReferenceDocx,
getReferenceODT have been removed from Shared and
moved into Class. They are now defined in terms of
PandocMonad primitives, rather than being primitve
methods of the class.
* toLang has been moved from BCP47 to Class.
* NoTranslation and CouldNotLoudTranslations have
been added to LogMessage.
* New module, Text.Pandoc.Translations, exporting
Term, Translations, readTranslations.
* New functions in Class: translateTerm, setTranslations.
Note that nothing is loaded from data files until
translateTerm is used; setTranslation just sets the
language to be used.
* Added two translation data files in data/translations.
* LaTeX reader: Support `\setmainlanguage` or `\setdefaultlanguage`
(polyglossia) and `\figurename`.
