Add option to specify table env. in LaTeX output #4392

pajowu · 2018-02-22T17:12:30Z

This PR adds a command line switch (--latex-table-environment) for switching between possible table environments. Currently pandoc is hardcoded to use longtable, which allows for multi-page-tables. This adds support for the tabularx environment (allowing columns that span over the while remaining table-width) and gives the possibility to easily add new environments in the future.

NOTE: The package used in the template could also be ltablex, which is a combination of longtable and tabularx (allows multi-page tables with auto-spanning columns).

pajowu · 2018-04-05T09:36:01Z

Is there something blocking this PR?

jgm · 2018-04-05T16:31:06Z

MANUAL.txt

+`--latex-table-environment=longtable`|`tabularx`
+
+:   Use the specified table environment when producing LaTeX output.
+    The default is [`longtable`] which allows table to spread over multiple tables.


Do you mean "over multiple pages"?

jgm · 2018-04-05T16:31:24Z

MANUAL.txt

+
+:   Use the specified table environment when producing LaTeX output.
+    The default is [`longtable`] which allows table to spread over multiple tables.
+    [`tabularx`] allows for columns which autospread over the entire linewidth.


I wasn't sure exactly what was meant by "autospread over the entire linewidth."

Maybe taking from the description of the ctan package:
" [tabularx] allows for paragraph-like columns whose width automatically expand to fill the width of the environment"

jgm · 2018-04-05T16:32:11Z

Sorry, I just haven't had time to review it. Can you say more about the motivation for the change?

themightyoarfish · 2018-10-19T12:48:35Z

One motivation from my perspective is that longtable is broken in the presence of floats on the same page (see e.g.here or here)

In my case, this happily pushes sections into the bottom margin and the space between pages, so they are lost.

This is a known bug in longtable, but unfortunately it also breaks pandoc use in cases, so an option for this would be great.

themightyoarfish · 2018-10-22T12:16:34Z

Could this be moved along? I don't have a workaround for this right now.

This commit adds a command line switch (`--latex-table-environment`) for switching between possible table environments. Currently pandoc is hardcoded to use `longtable`, which allows for multi-page-tables. This commit adds support for the tabularx environment (allowing columns that span over the while remaining table-width) and gives the possibility to easily add new environments in the future.

pajowu · 2018-11-09T11:20:48Z

I updated the documentation according to your Remarks and fixed all merge conflicts. The PR should be mergable again

Regarding my motivation: longtable inserts empty pages or overflows the page when float objects are on the same page. This makes longtable unusable for us. Longtable also does not have an option to say that one column should fill the rest of the linewidth, which we also need. With this patch applied and ltablex used as the table class, both issues are fixed

jgm · 2018-11-09T18:41:28Z

MANUAL.txt

+
+:   Use the specified table environment when producing LaTeX output.
+    The default is [`longtable`] which allows table to spread over multiple pages.
+    [`tabularx`] allows for paragraph-like columns whose width automatically expand to fill the width of the environment.


Lines should be hard-wrapped to fit within 80 columns.

I'm not sure I quite understand this description: "paragraph-like columns whose width automatically expand to fill the width of the environment." Maybe better to say something about what should dictate this choice? For example: If you have long tables that may need to spans multiple pages, use longtable. If you want to create floating tables (and for better compatibility with other floating elements on the page), use tabularx.

jgm · 2018-11-09T18:46:53Z

MANUAL.txt

@@ -1599,6 +1607,10 @@ LaTeX variables are used when [creating a PDF].
    supports 'plain' (default), 'empty', and 'headings'; headings puts
    section titles in the header.

+`table_environment`
+:   The environemnt given by `--latex-table-environment`


"environment"

jgm · 2018-11-09T18:48:16Z

src/Text/Pandoc/App/CommandLineOptions.hs

@@ -850,6 +850,17 @@ options =
                     UTF8.hPutStr stdout (usageMessage prg options)
                     exitSuccess ))
                 "" -- "Show help"
+    , Option "" ["latex-table-class"]


Shouldn't this be latex-table-environment, given the documentation? (also elsewhere)

jgm · 2018-11-09T18:49:46Z

src/Text/Pandoc/Options.hs

@@ -168,6 +169,11 @@ data ReferenceLocation = EndOfBlock    -- ^ End of block
                       | EndOfDocument -- ^ at end of document
                       deriving (Show, Read, Eq, Data, Typeable, Generic)

+-- | Class of tables to use in LaTeX output
+data LatexTableEnvironment = Longtable    -- ^ Use longtable for tables
+                     | Tabularx     -- ^ Use tabularx tables


This should align as in the other examples in the code.

jgm · 2018-11-09T18:50:29Z

src/Text/Pandoc/Options.hs

@@ -204,7 +210,8 @@ data WriterOptions = WriterOptions
  , writerReferenceLocation :: ReferenceLocation    -- ^ Location of footnotes and references for writing markdown
  , writerSyntaxMap         :: SyntaxMap
  , writerPreferAscii       :: Bool           -- ^ Prefer ASCII representations of characters when possible
-  } deriving (Show, Data, Typeable, Generic)
+  , writerLatexTableEnvironment   :: LatexTableEnvironment


Note clearly the API change in the commit comment.

jgm · 2018-11-09T18:52:50Z

src/Text/Pandoc/Writers/LaTeX.hs

  case align of
         AlignLeft    -> "l"
         AlignRight   -> "r"
         AlignCenter  -> "c"
-         AlignDefault -> "l"
+         AlignDefault -> if index == 0 then "l" else "X"


Why is "l" used for the first column and then "X"?

jgm · 2018-11-09T18:54:46Z

Thanks for the update. I've made some comments to your patch.

Currently there are no tests. There should be tests.

Also, your patch assumes that the only differences between the environments lie in the column specs and environment name. Have you tested things like captions, footnotes in table cells, and so on, to make sure that this is the case?

randrej · 2019-01-20T12:22:46Z

Wouldn't it make sense to use tabu instead of
tabularx?

tabu's environment longtabu mimics the longtable environment that it can
span multiple pages, but it allows for using x columns (columns with width
expanded so that the table takes up the whole \textwidth). It can also be used
just to stretch the table to full width by widening every column equally, just
like tabularx.

tabu also painlessly mends many of the concerns expressed here:

Also, your patch assumes that the only differences between the environments
lie in the column specs and environment name. Have you tested things like
captions, footnotes in table cells, and so on, to make sure that this is the
case?

Tabu is very much compatible with longtable as far as the table contents go.
While tabularx doesn't allow footnotes and verbatim text, tabu does. It can
also contain other tabu, tabular*, tabularx environments.

Like tabularx, tabu allows for specifying which columns are stretched in
the final output. Pandoc markdown currently doesn't have any way of specifying
that a column should be stretched, though. tabu can be used without
specifying which column to stretch. This stretches all of the columns.
Implementing this wouldn't require any changes to Pandoc markdown syntax.

New syntax spitballing

If you wanted to add syntax for column stretching, I have some suggestions.

I guess a solution that relies on the number of leading or trailing dashes is
out of question, since it could break someone's overly inset tables.

Hence, adding another character to the ---- lines to designate that the line
stretches seems like the simplest solution. I'll use the * character in my
examples, but it could be any character. A tilde ~ may be a nice choice, it's
visually similar to a dash. Adding a new character makes this effortlessly
backwards compatible.

Style 1:

* is placed on the side that specifies the alignment, and on both sides for
central alignment.

 def. aligned reg.  RA reg. str. LA   str. RA  str. CA  reg. LA
 ----------------- -------- -------* *------- *-------* ----------
 1                        2 3               4        5  6
 7                        8 9              10       11  12

Style 2:

* is always placed on the right side and is treated as a regular - for
alignment purposes

 def. aligned reg.  RA reg. str. LA   str. RA  str. CA  reg. LA
 ----------------- -------- -------* -------* --------* ----------
 1                        2 3               4        5  6
 7                        8 9              10       11  12

Similar syntax could be used for other types of tables. I could try to think
something up if you like this.

New syntax or no, tabu looks like a promising replacement for longtable. We
could add a global --latex-use-longtable option to ignore this syntax and
retain the things the way they currently are.

I believe that this extension to the syntax would bring value to users
(especially those who write scientific papers and anything else for
publishing). A writer could be easily written for HTML-based formats, too.

jgm · 2019-01-22T00:03:12Z

tabu scared me off with this note in the documentation:

…

The tabu package patches internal commands of many packages and is liable to stop working if those internal interfaces change. Unfortunately the original author appears to be out of contact and the package had not been updated for some years.

Grandement · 2019-04-27T12:23:56Z

I would prefer longtabu as well instead of longtable. The definition of longtabu is close to longtable and I guess it is not a big deal to add it with a optional argument to pandoc. The tabu packge is still working well :). With longtabu automatic linebreaks in a cell are possible due to the column X definition which would be a benifit for pandoc.

jgm reviewed Apr 5, 2018

View reviewed changes

link2xt added the format:LaTeX label Oct 25, 2018

pajowu force-pushed the ken/latex-table-environment branch from 2941904 to 01b9275 Compare November 9, 2018 10:11

pajowu force-pushed the ken/latex-table-environment branch from 01b9275 to f16e43e Compare November 9, 2018 11:16

jgm reviewed Nov 9, 2018

View reviewed changes

jgm mentioned this pull request Jun 6, 2019

longtable bug cause overflow #5557

Closed

jgm mentioned this pull request Jan 7, 2023

Text containing table flows beyond bottom page margin when figure floats to next page #8526

Open

jgm force-pushed the main branch from 3ff001b to 3da4d41 Compare December 18, 2024 23:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add option to specify table env. in LaTeX output #4392

Add option to specify table env. in LaTeX output #4392

pajowu commented Feb 22, 2018

pajowu commented Apr 5, 2018

jgm Apr 5, 2018

jgm Apr 5, 2018

agusmba Oct 22, 2018 •

edited

Loading

jgm commented Apr 5, 2018

themightyoarfish commented Oct 19, 2018

themightyoarfish commented Oct 22, 2018

pajowu commented Nov 9, 2018 •

edited

Loading

jgm Nov 9, 2018

jgm Nov 9, 2018

jgm Nov 9, 2018

jgm Nov 9, 2018 •

edited

Loading

jgm Nov 9, 2018

jgm Nov 9, 2018

jgm Nov 9, 2018

jgm commented Nov 9, 2018

randrej commented Jan 20, 2019

jgm commented Jan 22, 2019 via email

Grandement commented Apr 27, 2019

Add option to specify table env. in LaTeX output #4392

Are you sure you want to change the base?

Add option to specify table env. in LaTeX output #4392

Conversation

pajowu commented Feb 22, 2018

pajowu commented Apr 5, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

agusmba Oct 22, 2018 • edited Loading

Choose a reason for hiding this comment

jgm commented Apr 5, 2018

themightyoarfish commented Oct 19, 2018

themightyoarfish commented Oct 22, 2018

pajowu commented Nov 9, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jgm Nov 9, 2018 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jgm commented Nov 9, 2018

randrej commented Jan 20, 2019

New syntax spitballing

Style 1:

Style 2:

jgm commented Jan 22, 2019 via email

Grandement commented Apr 27, 2019

agusmba Oct 22, 2018 •

edited

Loading

pajowu commented Nov 9, 2018 •

edited

Loading

jgm Nov 9, 2018 •

edited

Loading