wrap new att lists in <ul> (where they were not so wrapped)

[sydb]

• template showAttClasses is called from 6 different places (all in common_tagdocs);
• generated debugging code for each of those places;
• found which 2 were causing problem (li element as child of td instead of child of ul);
• wrapped them in ul element.
• Note: deliberately left debugging PIs in place, but they should maybe be removed before merging into dev.

[ebeshero]

I'm looking for any trouble with @sydb 's repair in the output Guidelines. I see no problems with specs in:
[x] - Epub (inspected several element specs)
[x] - Guidelines PDF build
[x] - Guidelines HTML

Locally, the following tests pass:
[x] - TEI repo (no diffs from expected results)

Stylesheets Test 1 does not pass, breaks on

make[1]: *** [Makefile:583: test-oddity] Error 1
make[1]: Leaving directory '/stylesheet/Test'
make: *** [Makefile:106: test] Error 2

Stylesheets Test 2 passes on my Mac.
It's failing for me ONLY in Docker (and I suspect a Docker error?) on

• build.xml: 550
• build_docx.xml: 32, 43, and 64

[ebeshero]

Summarizing how things went by the end of the day on Saturday 10/29 (EST). @sydb notes (from Slack):

Every test in Test/ now passes except test-epub, for which 13 of the output files get the error “This file should declare in the OPF the property: scripted”.

In order to get the XSLT to run, the top of tei-to-epub3.xsl looks like:

  <xsl:import href="../html/html.xsl"/>
  <xsl:import href="../odds/teiodds.xsl"/>   <!-- only included for $autoGlobal param, which ../odds/classatts.xsl relies on -->
  <xsl:import href="../odds/classatts.xsl"/> <!-- only included for ATTCLASSES key -->
  <xsl:import href="../epub/epub-common.xsl"/>
  <xsl:import href="../epub/epub-preflight.xsl"/>
  <xsl:import href="../html/html_oddprocessing.xsl"/> <!-- only included for schemaOut template, which ../odds/teiodds.xsl seems to need -->

Syd notes: We need to figure out the correct file for the "schemaOut" template (there are 5 of them, all different).
All that work (i.e., importing 3 files which have a total of 4,390 elements and attributes) is being done just so we can execute this line in common_tagdocs.xsl :
<xsl:variable name="thisClassSpec" select="key('ATTCLASSES', @key)" as="element(tei:classSpec)?"/>

Perhaps this is better than using key():

<xsl:variable name="key" select="@key"/>
<!-- note: following line does not use "key('ATTCLASSES', @key)" because ATTCLASSES is not defined in this or any imported file, and importing it requires importing other files, too. -->
<xsl:variable name="thisClassSpec" select="//tei:classSpec[ @ident eq $key]" as="element(tei:classSpec)?"/>

[joeytakeda]

I've now tested this (adding the all of those <xsl:import>s as described above) and get the same validation problem with the epub:

Validating against EPUB version 3.0
ERROR: actual-results/test_3.epub/OPS/index.html: This file should declare in the OPF the property: scripted
ERROR: actual-results/test_3.epub/OPS/index-front.1_div.1.html: This file should declare in the OPF the property: scripted
ERROR: actual-results/test_3.epub/OPS/index-front.1_div.2.html: This file should declare in the OPF the property: scripted
[... plus many more ....]

I believe this is caused by an import, which is adding all of the HTML javascript into the epubs but shouldn't be there — or, at least, isn't there in test_3.epub from the last Stylesheets release's actual-results folder .

Adding the following to the bottom of epub3/tei-to-epub3.xsl allows make test-epub to succeed:

  <!--Add empty javascript hook to remove JS-->
  <xsl:template name="javascriptHook"/>

The suggestion of replacing key() with a variable (and not adding a bunch of imports) is probably the right call; however, when I did that, different parts of Test failed, so some more investigation would be needed there to see if there are different and unexpected repercussions (sigh)

[sydb]

Implemented change I suggested above and which @joeytakeda thought was the way to go in 7ed5111. Has passed test of building the Guidelines. Running Stylesheet tests now.

[sydb]

After removing <xsl:import href="../odds/classatts.xsl"/> from epub3/tei-to-epub3.xsl (which I forgot to do in previous commit, now done in 05231d4), this branch passes Test2/ but fails Test/ with a diff error (albeit one we may eventually to decide is OK).[1]

In Test/ the file actual-results/test15.odd.html has 4 extraneous empty <li class="classSpecItem" itemprop="item"> elements. Should we be worrying about that?

Note
[1] This is a bit troublesome because it means Test2/ is missing something. But that is not the subject of this PR.

[joeytakeda]

In Test/ the file actual-results/test15.odd.html has 4 extraneous empty <li class="classSpecItem" itemprop="item"> elements. Should we be worrying about that?

@sydb: Do we know what is it that's causing them? At the moment, it produces an output that has (unsurprisingly) empty additional list items (below screenshot of output of actual-results/test15.odd.html), so I think we do need to worry about that:

I wonder if we should reverse course and take the brute-force approach of adding the imports and the empty javascript hook? While this is certainly more overhead, it does allow both test suites to pass (and I don't imagine would cause problems in other transformations, unlike changes to common)

[sydb]

No, I do not really know what is causing them.

As for going the “adding the imports and the empty javascript hook” route, do we know what this test file looks like with that system?

[sydb]

I found the bug, I think: the //classSpec that I used instead of the ATTCLASSES key needed to look at only <classSpec>s with an @type of "atts". We added the needed predicate. @martindholmes is pushing up the results now.

[martindholmes]

This appears to be working. Ready for merge as soon as the tests all complete.