HCMC Blogs

Struggling with the strange behaviour I was seeing, where a function would execute correctly if called directly, but not if called from a for loop, I discovered two things: the same problem still exists in eXist 1.4.1, but there I see an actual error to the effect that the context is missing for a node; and I can eliminate the problem by rewriting some XPath inside the query. This is the XPath that was causing the problem:

for $n in (distinct-values($doc//@key[parent::tei:name[not(@type)]]))

Admittedly it's a bit perverse. If it's rewritten like this:

for $n in (distinct-values($doc//tei:name[not(@type)]/@key))

then the query works even when executed inside a loop. This means I can now generate all the complete OAI records that JD would like to see.

I have added Page-Break tags to the CO 60/5 files. All that remains for the remaining 1859 files is CO 60/6, the images for which Theo will be done with by week's end. In the meantime, I can begin tagging people, place, ship, and mentions of First Nations in the 1859 files, generally.

Left early on Friday.

I've hit a snag with the OAI stuff which is almost certainly an eXist bug: when I generate an OAI record individually, passing an xml:id to the function that does it, then all the people, places etc. are included, but when I try to generate records in a loop by passing all the ids in, then that information is not included. I've tried configuring extra indexes and all sorts of other workarounds but it seems insuperable, and I'm certainly not going to generate 7,143 records one at a time. I'm now forced to look at updating to the new build of eXist/Cocoon to see if the bug is present there. If it's not, then that's a solid reason for doing the migration right now.

Moving forward on this proposal -- should be done early next week.

Jamie and I met with CC and PS to hammer out a priority list of tasks to complete before CALICO.

By April 20th PS will have
1) fixed 2 or 3 small interface bugs we ID'd
2) completed a basic shell for the map (just a GUI wrapper)
3) tweaked the search results page so it produces thumbnailed links (timestamped) in to each video
4) convert new pages to HTML5
5) embed an HTML5 video player (with Jamie's help)

By April 20th HCMC will produce
1) a basic map displaying all locations of speakers currently in the db, with each location providing a hook in to the search system to provide a list of speakers from that location. It will use OpenLayers and KML to begin with.
2) The "Other videos" panel to the right of the video player panel will be dynamic, providing, at first, a selection of other speakers from the same region as the speaker currently in the video player panel. Other possible categorizations were discussed, but not finalized.

After April 20th we will discuss any leftover issues and decide how to proceed. The goal is to present a mostly feature-complete version at CALICO (May 17-21).

A further 1250 page-images have been added to the manuscript image browser, covering British North America 1859 Public Offices/Hudson's Bay Company. Transcriptions are now being linked into these images.

Met with ML re Zotero and a group bibliography. We're going to run a test with three of us, performing the roles of submitter, admin, and non-member, to see if the desired workflow can be achieved inside Zotero.

There were some oddities in the display of references linked from inside references -- see my prior post for details. I've now fixed those. The problems arose because:

1. When a reference was linked from inside another reference, it wasn't necessarily already in the page, because only references linked directly from the text are pulled into the document initially. That meant it had to be retrieved by AJAX.
2. When it was retrieved by AJAX, it was a fragment rather than a full document; that meant that a lot of the context which was used by XSL matching templates was absent, so a lot of the styling was not applied.
3. When the reference was retrieved, it was supposed to be added to the list of existing references. However, it replaced them instead, meaning that all subsequent reference links had to retrieve their references anew -- and they then inherited the style problems described above.

I've now fixed this so that the transformation XSL is more sensitive to fragments (it uses a formula local-name(/*) ne 'TEI' to identify a fragment, which is typically rooted on <div> rather than <TEI>). I've also tweaked the AJAX code so that it can append to the content of the target element as well as replace it, using a switch called clearTargetFirst.