The process was too fragile. It made too many assumptions about
available libraries (which failed sometimes when sandboxes were
used). This is a low-tech solution. The only drawback is that
`man/pandoc.1` is a generated file in the repository. It will need
to be regenerated periodically when README changes.
The pandoc.1 man page is generated automatically after the cabal
build process. It goes in `data/pandoc.1`. It can be obtained
by the user who installs pandoc via cabal thus:
pandoc --print-default-data-file pandoc.1 > pandoc.1
+ Removed `--man1`, `--man5` options (breaking change).
+ Removed `Text.Pandoc.ManPages` module (breaking API change).
+ Version bump to 1.15 because of the breaking changes, even
though they involve features that have only been in pandoc
for a day.
+ Makefile target for `man/man1/pandoc.1`. This uses pandoc to
create the man page from README using a custom template and filters.
+ Added `man/` directory with template and filters needed to build
man page.
+ We no longer have two man pages: pandoc.1 and pandoc_markdown.5.
Now there is just pandoc.1, which has all the content from README.
This change was needed because of the extensive cross-references
between parts of the README.
+ Removed old `data/pandoc.1.template` and
`data/pandoc_markdown.5.template`.
This change adds `--man1` and `--man5` options to pandoc, so
pandoc can generate its own man pages.
It removes the old overly complex method of building a separate
executable (but not installing it) just to create the man pages.
The man pages are no longer automatically created in the build
process.
The man/ directory has been removed. The man page templates
have been moved to data/.
New unexported module: Text.Pandoc.ManPages.
Text.Pandoc.Data now exports readmeFile, and `readDataFile`
knows how to find README.
Closes#2190.
In 1b44acf0c5 we replaced some
hackish CSS parsing with css-text, which I thought was a complete
CSS parser. It turns out that it is very buggy, which results
in lots of things being silently dropped from CSS when
`--self-contained` is used (#2224).
This commit replaces the use of css-text with a small but
more principled css preprocessor, which only removes whitespace
and replaces URLs with base 64 data when possible.
Closes#2224.
* Reverted kludgy change to make-windows-installer.bat.
* Removed make-reference-fiels.hs.
* Moved the individual ingredients of reference.docx and
reference.odt to the data directory.
* Removed reference.docx and reference.odt from data directory.
* We now build the reference archives from their ingredient pieces
in the docx and odt writers, instead of having a reference.docx
or reference.odt intermediary.
This should fix#2187.
It also simplifies the bulid procedure.
The one thing users may notice is different is that you can
no longer get the reference.docx or reference.odt using
`--print-default-data-file`. Instead, simply generate a
docx or odt using pandoc with a blank or minimal input,
and use that (or a customized version) with `--reference-docx`
or `--reference-odt`.
We only support the href attribute, as there's no place for
"target" in the Pandoc document model for links.
Added HTML reader test module, with tests for this feature.
Closes#1751.
This ensures that all code blocks will be wrapped in a div
with class sourceCode. Also, the default highlighting CSS
now adds `div.sourceCode { x-overflow: auto; }`, which means
that code blocks (even with line numbers) will acquire a scroll
bar on screens too small to display them (e.g. mobile phones).
See #1903 and jgm/highlighting-kate#65.
- Added commonmark as an input format.
- Added `Text.Pandoc.Readers.CommonMark.readCommonMark`.
- For now, we use the markdown writer to generate benchmark
text for the CommonMark reader. We can change this when we
get a writer.
This allows the test suite to be run using "+RTS -N".
Doing so improves the performance of the test suite on my quad-core Mac laptop as follows:
Before: 8.2 seconds
After: 2.5 seconds
Build dependencies of the trypandoc executable are required, regardless of the
trypandoc flag was set to either True or False. Correct package description
to make them truly optional.
Renamed some tests, introducing subsidiary directories
for fb2, docx, epub.
Cleaned up tests in cabal file.
Combined dokuwiki-writer and dokuwiki_inline_formatting tests.
This was just too fragile and dependent on a changing Cabal API
(see #1526).
Instead of passing the bulid directory to the test program, we
now let the test program find itself (using executable-path)
and then find the pandoc executable relative to itself.
Rewrote features test to remove all unimplemented features.
There are now all three examples of where an image can be included in
the test.
1. Cover image
2. As a spine elemnt
3. In the document
Tests have also been added to make sure that the mediabag contains all
these images after processing.
Renamed epub test files so they're identified more clearly as
epub: features.{epub,native} -> epub.features.{epub,native},
and similarly with formatting.{epub,native}.
Added epub test files to cabal file, so they'll be included in
the tarball.
This will allow us to test the whole mediabag (making sure, for example,
that images are added with the correct keys) instead of just individual
extracted images. We compare each entry in the media bag to an image
extracted on the fly from the docx. As a result, we only need one file
to test with.
The image in the current tests was also replaced with a smaller one.
Moved `MediaBag` definition and functions from Shared:
`lookupMedia`, `mediaDirectory`, `insertMedia`, `extractMediaBag`.
Removed `emptyMediaBag`; use `mempty` instead, since `MediaBag`
is a Monoid.
http://txt2tags.org/
There are two points which currently do not match the official
implementation.
1. In the official implementation lists can not be nested like the
following but the reader would interpret this as a bullet list with the
first item being a numbered list.
```
- + This is not a list
```
2. The specification describes how URIs automatically becomes links.
Unfortunately as is often the case, their definitiong of URI is not
clear. I tried three solutions but was unsure about which to adopt.
* Using isURI from Network.URI, this matches far too many strings and is
therefore unsuitable
* Using uri from Text.Pandoc.Shared, this doesn't match all strings that
the reference implementation matches
* Try to simulate the regex which is used in the native code
I went with the third approach but it is not perfect, for example
trailing punctuation is captured in Urls.
mtl switched from ErrorT to ExceptT, but we're not sure which mtl we'll
be dealing with. This should make errors work with both.
The main difference (beside the name of the module and the monad
transformer) is that Except doesn't require an instance of an Error
Typeclass. So we define that for compatability. When we switch to a
later mtl, using Control.Monad.Exception, we can just erase the instance
declaration, and all should work fine.
With the move from parsec to attoparsec, we lost good error
reporting. In fact, since we weren't testing for end of input,
malformed templates would fail silently. Here we revert back to
Parsec for better error messages.
This has fixes for unicode path names. Note that compiling
pandoc against zip-archive 0.2.3 or 0.2.3.1 will lead to invalid
zip containers, causing LibreOffice (e.g.) to regard ODTs as corrupt.
They aren't needed at runtime.
We keep README and COPYRIGHT in data to ensure that they'll be
available on all systems on which pandoc is installed.
Closes#1123.
Users of s5 and slideous will have to download the needed
files, as has been documented for some time in the README.
slidy files will be sought on the web, as before.
Previously we tried to remove make-pandoc-man-pages from the list
of packages to be haddocked, installed, copied, etc.
It works better to set 'Buildable: False' on make-pandoc-man-pages,
then have the buildHook temporarily set Buildable to True. This
allows make-pandoc-man-pages to be built (and used in generating
the man pages), but not installed.
Note that anything not parseable as a YAML boolean or string
is treated as a literal string.
Note that you can still get a string value with "yes" or any
of the strings interpretable as booleans:
-M boolvalue=yes -M stringvalue='"yes"'
Going forward we'll use pandoc-citeproc, as an external filter.
The `--bibliography`, `--csl`, and `--citation-abbreviation` fields
have been removed. Instead one must include `bibliography`, `csl`,
or `csl-abbrevs` fields in the document's YAML metadata. The filter
can then be used as follows:
pandoc --filter pandoc-citeproc
The `Text.Pandoc.Biblio` module has been removed. Henceforth,
`Text.CSL.Pandoc` from pandoc-citations can be used by library users.
The Markdown and LaTeX readers now longer format bibliographies and
citations. That must be done using `processCites` or `processCites'`
from Text.CSL.Pandoc.
All bibliography-related fields have been removed from `ReaderOptions`
and `WriterOptions`: `writerBiblioFiles`, `readerReferences`,
`readerCitationStyle`.
API change.
This way filters can figure out what the target format is
and react appropriately.
Example:
#!/usr/bin/env runghc
import Text.Pandoc.JSON
import Data.Char
main = toJSONFilter cap
where cap (Just "html") (Str xs) = Str $ map toUpper xs
cap _ x = x
This capitalizes text only for html output.
* Text.Pandoc.Writers.Shared contains shared functions used
only in writers.
* metaToJSON now takes a WriterOptions parameter, and will
return an empty object if standalone is not specified.
* Depend on pandoc 1.12.
* Added yaml dependency.
* `Text.Pandoc.XML`: Removed `stripTags`. (API change.)
* `Text.Pandoc.Shared`: Added `metaToJSON`.
This will be used in writers to create a JSON object for use
in the templates from the pandoc metadata.
* Revised readers and writers to use the new Meta type.
* `Text.Pandoc.Options`: Added `Ext_yaml_title_block`.
* Markdown reader: Added support for YAML metadata block.
Note that it must come at the beginning of the document.
* `Text.Pandoc.Parsing.ParserState`: Replace `stateTitle`,
`stateAuthors`, `stateDate` with `stateMeta`.
* RST reader: Improved metadata.
Treat initial field list as metadata when standalone specified.
Previously ALL fields "title", "author", "date" in field lists
were treated as metadata, even if not at the beginning.
Use `subtitle` metadata field for subtitle.
* `Text.Pandoc.Templates`: Export `renderTemplate'` that takes a string
instead of a compiled template..
* OPML template: Use 'for' loop for authors.
* Org template: '#+TITLE:' is inserted before the title.
Previously the writer did this.
Currently the library is set up with a shim for association
lists, for compatibility, but this can change when the writers
are changed.
New export: `varListToJSON`.
Removed `Empty`.
Simplified template type to a newtype.
Reason: the path to the system temp directory may contain tildes,
which causes problems in LaTeX when the username is more than
eight characters.
Closes#777.
pandoc -t data/sample.lua
will load the script sample.lua and use it as a custom writer.
data/sample.lua is provided as an example.
Added `--print-custom-lua-writer` option to print the sample
script.
The _note attribute is supported. This is unofficial, but
used e.g. in OmniOutliner and supported by multimarkdown.
We treat the contents as markdown blocks under a section
header.
Added to documentation and tests.
