dialoa / collection Goto Github PK

Provide an option to continue example lists from one source to another

Without or with --extract-media.

A document can hyperlink to files of any type. When links have relative targets, we need to make sure that they're correct in the resulting document. What are the options?

1. Mimick pandoc's behaviour. Let the source files relative links as they are.
2. Adjust to make them relative to the master file directory.
3. Adjust to make them relative to the present working directory.
4. Adjust to make them relative to the output file directory.

3 is needed to make image inclusion in PDF/epub and extract-media work. We should cater to 1 in case the user is already entering the links with a specific root in mind. We might want to cater for 2 as it mimicks having writen a big master document directly with links relative to it. We might want to cater for 4 to ensure outputs work out

I suggest implementing 3 and 1. Users can achieve 2 and 4 by running pandoc from the master or output directory, respectively and relying on 3.

When gathering bibliography fields from sources, adjust relative paths. On by default, option to deactivate.

Currently in native import mode we run pandoc -t json. This is useless if people use filters that need to have the ultimate output format in FORMAT.

Solution: we run pandoc -t FORMAT (adjusting html when building an epub), but we catch the AST object rather than the output. To do this we need a filter that outputs the AST as json in a temp file that we can pick up. The command would look something like this:

pandoc source -t FORMAT -L catch.lua -M catch-filter-output=temporary.json

When targetting LaTeX pandoc-crossref (and possibly other filters) turn citations in to \ref commands. Since we run isolate after other filters, these aren't prefixed.

• one solution would be run isolate before and recognize the pandoc-crossref pseudo-citation links, revise them. Pb: doesn't work for other filters, you still need to run isolate after other filters in case they created links.
• another solution is for isolate to scan the rawinlines too for LaTeX ref commands.

NB, check out the case in which the RawInline includes more than \ref, e.g. some formatting (\ref{...}), or perhaps when the ref is within a large LaTeX block.

Pandoc sets some template variables automatically: $math$ if there's math, $csl-refs$ if there are citeproc references, and so on. In Raw mode these are set wrong because by the time the raw blocks are imported in the master document Pandoc doesn't see the citeproc references and the maths.

Workaround: the template should include needed LaTeX code unconditionally. Downside: will load mathjax only if not needed.

Workaround 2: make our own duplicates of Pandoc's internal template variables, and pass them from sources to main?

See example 04, recursive mode. Currently the sub-master file needs to specify the location of its children relative to the main master.

If a user runs the collection with --extract-media, we need to extract the media from every source and put them in a common folder. Provided each source's path in the combined document is relative to the pwd, Pandoc will reproduce the relevant folder structure in the new media bag (at least if the locations are subfolders of pwd), so this alone should prevent any filename conflict. Check that this is the case.

Citeproc crashes when the bibliography field is present but empty. It would be nice for collection to be able to validate import sources metadata before building them.

Currently the filter simply throws metadata at the imported file with --metadata-file flags, without e.g. first integrating the generic and child metadata specified in master. In a redesign it would first build the metadata first, validate it, and then send it to the imported file.