BUG: Fix Table 2.1 exceeding the page width.

[jhlegarreta]

Fix Table 2.1 in chapter CONFIGURING AND BUILDING ITK exceeding the
page width.

Fixes #62.

[jhlegarreta]

Not sure if the rendered table will be too small.

If it is, an alternative would be to use a landscape layout on a page of its own. Looks like there are a number of alternatives:

• Using the rotating or the afterpage packages:
  https://tex.stackexchange.com/questions/19017/how-to-place-a-table-on-a-new-page-with-landscape-orientation-without-clearing-t

• Using the lscape package:
  https://tex.stackexchange.com/questions/345483/horizontal-page-for-largetables/345484

• Using the pdflscape package:
  https://texblog.org/2007/11/10/landscape-in-latex/

I've tested rotating, pdflscape and lscape in some separete, MWEs and all seem to work well. rotating and lscape just rotate the content, whereas pdflscape rotates the page itself as well, which I ignore whether it is desirable. Finally, rotating requires using the command \begin{sidewaystable} whereas lscape does not (just using \begin{landscape} \end{landscape} around the table). Either are fine for me if the rest of the pages before/after render OK.

[thewtex]

I think we have to just rotate the content, not the page, so that it can continue to be printed as a book.

[jhlegarreta]

Thanks for the review @thewtex ! Commit 5bf45e4 implements a landscape orientation for the table while keeping the portrait orientation for the page. Let's check whether the output provided by the CI builds is according to the expectations.

[thewtex]

👍 Cool, @jhlegarreta !

The build is failing. Do we need \usepackage{rotating}?

[jhlegarreta]

Yes, sorry for the oversight @thewtex. a212164 addresses this. Thanks for the review.

[jhlegarreta]

Thanks @thewtex. Nonetheless, I've checked the output and it does not quite look like we expected. I need to investigate further.

[jhlegarreta]

Included the legend in the sidewaystable environment, which I guess we want. Also, since the table is not immediately below the cross-reference, used colon instead of period.

The table overlapping the header is still unsolved. I've been looking for a solution without success. Will try to look better.

[jhlegarreta]

One option that I do not particularly like (since it is a workaround rather than a solution) is to change the font size in the table to small. Another one is customize the header for float-only pages, as suggested in this post. But again, that may have undesired effects on pages that may be difficult to spot in a long document like ours.

[jhlegarreta]

I do not see an explanation to the CI build failure. @thewtex how can I re-run the CI build?

[thewtex]

@jhlegarreta When you click the Details link next to ci/circleci: build, there should be a button, Rerun job with SSH. You can click that, and it will restart the job but also provide a temporary ssh command to log into the system, inspect outputs, etc. If the button is not there, please let me know, and we will get you the required permissions.

[jhlegarreta]

@thewtex that's right: I had already navigated to the cicle CI build to inspect the output and tried Rerun job with SSH but it is not a clickable button for my user it appears. I get the You need write permissions to trigger builds message when I hover it.

[thewtex]

@jhlegarreta That's interesting -- maybe a CircleCI bug? You are on the ITKSoftwareGuide developers team, which has Admin privileges on the repository.

I added your account to the Collaborators for the repository with Write access. Please try again to see if that does it.

[jhlegarreta]

Sorry, @thewtex , still the same result.

[thewtex]

How about after pushing to the branch?

[jhlegarreta]

Will try tomorrow afternoon then. Thanks for the suggestion !

[jhlegarreta]

@thewtex I discovered why: I did not have my GitHub account connected in my circle CI ' Account Integration` settings. Now, I ignore whether just being part of the Admins would have been enough... Sorry Matt. Re-running the job now...

[thewtex]

Ah, that was tricky. Thanks for taking a look, @jhlegarreta

[jhlegarreta]

This is weird: checking out the code for the commit hash at issue (caa048f) is failing in circle CI:

fatal: Could not parse object 'caa048fe9d21116f734f4557fcd635a83215cd42'.

No idea why this should be happening.

[thewtex]

@jhlegarreta that is odd...

Since caa048fe9d21116f734f4557fcd635a83215cd42 is the commit on this branch, please try changing the content, then git commit --amend so a new commit is created, then force pushing.

[jhlegarreta]

@thewtex Sorry for the delay 😬 I've just rebased and pushed. Let's see what the CI builds tell, and if we're able to solve this.

[jhlegarreta]

Patch set c7bc943 still showed a table stepping into the header. An alternative to make the table smaller is

\usepackage{lscape}
....
\begin{landscape}
\begin{table}
\begin{center}
\resizebox{1.6\textheight}{!}{
\begin{tabular}{c c c c c c c c c c c c c c c c c}
\toprule
...
\end{tabular} }
\end{center}
...

But still the look of a lonely table in a single page did not look nice to me, and hence in f7fb0f6 I opted for coming back to a portrait orientation table but limiting its width. If we are happy with the result, I'll update the commit message and this will be ready to be merged.

[jhlegarreta]

The result looks satisfactory to me. Check it at
https://open.cdash.org/upload/c775d762a71fe6cf7f601d62263f8887f39d8140/ITKSoftwareGuide-Book1.pdf

It's just the position that LaTeX chose that makes the compiler list time schedule be broken that is not that nice. I've added a [h] in 4aaf5b3 to see if that forces LaTeX choose a better position and avoids breaking the list.

Updated the commit message.

[jhlegarreta]

Looked OK to me but for the legend: I forgot to force its position 😊. It should look OK in 960378f !

[jhlegarreta]

It does look OK to me 👌 now !

[thewtex]

Way to go, @jhlegarreta !!