OpenDocument writer: Table width support

[pyssling]

Use the same key-value in Attr for the table as the HTML writer
to implement table width support.

[tarleb]

Thanks! I left some minor code comments. LGTM otherwise!

[tarleb]

Just to keep you in the loop: a new release is in preparation right now. That means merging may be delayed until after the release is out.

[jgm]

Can you explain more fully what you mean by "table width support"?

[pyssling]

Yes sorry - this is the width of the table as percent of the page-width. It's correctly transferred into the HTML writer.

[jgm]

This should be determined by the relative widths of the columns, as specified in the table AST, right?

[pyssling]

Column widths come in as a percentage of the table width as far as I can tell. I.e. you still need the full width of the table.

The AST types define the following:

-- | The width of a table column, as a fraction of the total table
-- width.

I.e. they seem to be relative to the total table width - not absolute.

[jgm]

No, actually, the proper interpretation of the relative column widths in the table AST is as percentages of text width (not table width).

[pyssling]

Ok, I can change it to do that and send a pull-request to fix the AST documentation then.

[pyssling]

@jgm here is a pull-request to correct the AST documentation: jgm/pandoc-types#85

If you could review and it's correct then I can refactor this pull-request and the one for the DocBook reader.

[pyssling]

@jgm I've refactored this now to make use of the column widths as a percentage. Much more satisfying than looking at an undocumented key-value I have to say.

[jgm]

I'm just wondering how this interacts with the relative column widths. Do we support column widths yet in opendocument? If we do, how do these interact with the table width?

[pyssling]

Yes - column widths are supported and work from what I've seen - they are expressed in proportion to each other - i.e. they are without unit.
It works perfectly from what I can see - it really just scales the table in relation to text-width - doesn't affect the columns at all - they are still proportional as before.

[jgm]

How about a test or two? Easiest way is to add a command test, which could be named test/command/6792.md.

[pyssling]

I'll see what I can do - this requires a full test - command tests do not work for styles as they can only manage opendocument output - not odt output.

[pyssling]

I'm afraid this is beyond me - there is no framework for testing the ODT writer - just the ODT reader. I'm afraid I wouldn't know how (or have the time) to develop that sort of framework. I'm afraid these styles will remain untested for the time being.

[jgm]

Since this modifies the opendocument writer, couldn't we just have a command test with pandoc -t opendocument?

[pyssling]

Ok, thank you for pushing me on this. I figured out what was going on - the opendocument writer doesn't emit styles unless a template is used - so I got that working. Now there's a solid unit test. Thank you.

[jgm]

Great. I don't think we need the special template. Just use the default one:
pandoc -f native -t opendocument -s.

[pyssling]

done, sorry about that - didn't realize the template directory was accessible.

[jgm]

great, thanks!