setup typesense

[cfhaak]

As soon as #7 is done you could basically start creating the search index. But maybe provide some help with other tasks first … You could e.g. split the development of the search page between front (instant search) and backend (typesense index building)…

[cfhaak]

Anyway have a look at the roadmap first …

[fsanzl]

• [x] Search index
• [x] Volltext und Datum
• [x] nach Person fazetieren

[cfhaak]

for creating the fulltext use peters pyutil function https://github.com/acdh-oeaw/acdh-tei-pyutils/blob/517c0c8f7f6cd934fe7ec873da660ec9c763b47b/acdh_tei_pyutils/utils.py#L26

[fsanzl]

We need a xml:id for each document. Addressed in https://github.com/fun-with-editions/staribacher-data/issues/5

[cfhaak]

Nice! The search looks good so far!

[fsanzl]

• [x] Index persons
  • [x] use ref -> xml:id -> unique full name
  • [x] Fazettierung
• [x] Day/month/year
• [x] 'Date' instead of 'Title'

[fsanzl]

According to TS manuals, it is advisable to represent dates as Unix timestamps, so they can be directly tested with numeric operators <, >, =, !=. Some source files contain texts covering a range of dates while others have a single day. I have added two additional int32 objects to the TS schema to represent the date attributes @from and @to, which become the same value if the document provides just a @when.

[cfhaak]

Cool! That should work!

[fsanzl]

{
    'project': 'STB',
    'id': 'staribacher__19781220',
    'resolver': '/staribacher__19781220.html',
    'rec_id': 'staribacher__19781220.xml',
    'title': 'Mittwoch, der 20. Dezember 1978',
    'year': '1978-12-20',
    'notbefore': 282956400,
    'notafter': 282956400,
    'persons': [
        'Bacher, Gerd',
        'Benya, Anton',
        'Broda, Christian',
        'Burian, Ferdinand',
        'Frank, Wilhelm',
        'Frauscher, Reinhard',
        'Fremuth, Walter',
        'Gratz, Leopold',
        'Heindl, Kurt',
        'Jagoda, Karl',
        'Kienzl, Heinz',
        'Kreisky, Bruno',
        'Lachs, Thomas',
        'Marsch, Gerhard',
        'Riedl, B',
        'Schwarz, Walter',
        'Tischler, Margarete',
        'Wanke, Otto',
        'Wirlandner, Stefan',
        'Zilk, Helmut'],
    'anchor_link': 'p__35',
    'full_text': 'Der Energieplan wird spätestens Anfang des nächsten Jahres vorliegen. Nach Meinung Franks sollte er unbedingt als ein Bericht an den Nationalrat gesendet werden. Das Rohstoffkonzept wird durch die Wirtschaftsforschungsunterlagen schön langsam gestaltet.'}

[fsanzl]

xml:id becomes a random-looking html id through <p id="{local:makeId(.)}" class="yes-index> when transformed by editions.xsl instead of being kept as-is.

[csae8092]

xml:id becomes a random-looking html id through <p id="{local:makeId(.)}" class="yes-index> when transformed by editions.xsl instead of being kept as-is.

you can remove this xsl directive, or even better rewrite it so that it only applies on tei:p elements without @xml:ids

[fsanzl]

Sorry, I forgot the indication above. I already sorted the <p id> out yesterday

[fsanzl]

• [x] Persons does something but not quite what was expected
• [x] Months go 0-11 instead of 1-12 in slider.
• [x] Dates on top show the internal timestamp instead of presenting them in a human-readable guise
• [ ] CSS to address white chars on white bg

[fsanzl]

It seems ok to me. The font color of the refinements would be better addressed in the CSS.