Previously we had tested certain properties of the output PowerPoint
slides. Corruption, though, comes as the result of a numebr of
interrelated issues in the output pptx archive. This is a new
approach, which compares the output of the Powerpoint writer with
files that we know to (a) not be corrupt, and (b) to show the desired
output behavior (details below). This commit introduces three tests
using the new framework. More will follow.
The test procedure: given a native file and a pptx file, we generate a
pptx archive from the native file, and then test:
1. Whether the same files are in the two archives
2. Whether each of the contained xml files is the same. (We skip time
entries in `docProps/core.xml`, since these are derived from IO. We
just check to make sure that they're there in the same way in both
files.)
3. Whether each of the media files is the same.
Note that steps 2 and 3, though they compare multiple files, are one
test each, since the number of files depends on the input file (if
there is a failure, it will only report the first failed file
comparison in the test failure).
Closes#4281.
Previously we allowed "nonindent spaces" before the
opening and closing `:::`, but this interfered with
list parsing, so now we require the fences to be
flush with the margin of the containing block.
Now list item contents is parsed as blocks,
without resorting to parseFromString.
Only the first line of paragraph has to
be indented now, just like in Emacs Muse
and Text::Amuse.
Definition lists are not refactored yet.
See also: issue #3865.
Previously we could get ever-lengthening cell widths
when a table was run repeatedly through `pandoc -f markdown -t
markdown`. This patch stabilizes the relative cell
widths. Closes#4265.
This is difficult to recreate with a modern version of Word, so I'm
using the file submitted with the bug report. It would be preferable
to find a smaller example with Latin characters, though, so as not to
confuse the issue being tested.
rst2latex.py uses an align* environment for math in
`.. math::` blocks, so this math may contain line breaks.
If it does, we put the math in an `aligned` environment
to simulate rst2latex.py's behavior.
Closes#4254.
The change both improves performance and fixes a
regression whereby normal citations inside inline notes
were not parsed correctly.
Closesjgm/pandoc-citeproc#315.
even if the `latex_macros` extension is set.
This reverts to earlier behavior and is probably safer
on the whole, since some macros only modify things in
included packages, which pandoc's macro expansion can't
modify.
Closes#4246.
Elements with attributes got an additional `attr` accessor. Attributes
were accessible only via the `identifier`, `classes`, and `attributes`,
which was in conflict with the documentation, which indirectly states
that such elements have the an `attr` property.
Every constructor which accepts a list of blocks now also accepts a
single block element for convenience. Furthermore, strings are accepted as
shorthand for `{pandoc.Str "text"}` in constructors.
We had previously defaulted to slideLevel 2. Now we use the correct
behavior of defaulting to the highest level header followed by
content. We change an expected test result to match this behavior.
There isn't any reason to have numberous anchors in the same place,
since we can't maintain docx's non-nesting overlapping. So we reduce
to a single anchor, and have all links pointing to one of the
overlapping anchors point to that one. This changes the behavior from
commit e90c714c7 slightly (use the first anchor instead of the last)
so we change the expected test result.
Note that because this produces a state that has to be set after every
invocation of `parPartToInlines`, we make the main function into a
primed subfunction `parPartToInlines'`, and make `parPartToInlines` a
wrapper around that.
Previously we had only read the first child of an sdtContents tag. Now
we replace sdt with all children of the sdtContents tag.
This changes the expected test result of our nested_anchors test,
since now we read docx's generated TOCs.
