Tests were recently updated to use a "tail-buffer"-based lookup to account for text with multiple accession numbers and arbitrary text (aka a "wall label"):

• https://github.com/sfomuseum/accession-numbers/blob/main/cmd/test-runner/main.go#L98-L170

For example:

"This is an object\\nGift of Important Donor\\n302.2021.x1-x2\\n\\nThis is another object\\nAnonymouts Gift\\nPG731.2019 146.2020": 3

This breaks the AIC test for Obj: 96681 (which has been given an expected count of -1 so as not to stop other tests):

> make tests
env GOOS=darwin GOARCH=amd64 go build -o bin/darwin/test-runner cmd/test-runner/main.go
env GOOS=linux GOARCH=amd64 go build -o bin/linux/test-runner cmd/test-runner/main.go
env GOOS=windows GOARCH=amd64 go build -o bin/windows/test-runner cmd/test-runner/main.go
bin/darwin/test-runner data/*.json

2021/11/16 11:15:54 Failed to run tests for Art Institute of Chicago, Failed to find matches for 'Obj: 96681' using '((?:[RXTNObjf0-9: ]+)\/?(?:[0-9\.\-,]*)|(?:\d{4})\.(?:\d+[a-z\-]*))' (Art Institute of Chicago), expected 1 matches but got 2 ([Obj: 96681])
make: *** [tests] Error 1

@nikhiltri Any thoughts on the best approach here?

We'd like to contribute the V&A accession number patterns to the project. We've executed the test-runner on the json and it has passed all the tests.

Hi, I'm trying to add americanart.si.edu, but I'm having some issues getting the regex to pass the Linux test-runner. Confusingly, the test-runner says it fails on a different accession number each time it's run against the same pattern.

Museum of New Zealand Te Papa Tongarewa https://www.tepapa.govt.nz (([A-Z]+[0-9]+[/A-Z])|((?:\d+)-(?:\d+)-(?:\d++)(/[A-z\ -]+))) ME001533 2014-0031-1 GH004839 2008-0006-1/A-L to L-L OL000491 SP082549/A

Musée National d'Art Moderne (Centre Pompidou) https://www.mam.paris.fr (([A-Z]+)(\ |-|-)(?:\d+)((-[0-9]+)|(\ ([0-9]+))|([A-Z]+))*) AMPH 204 AMS 403 AMVP-2015-277 AMVP-2015-975 AME 1104 (2) AMPH 871TER

Write a line-separated text file containing names (URLs) of *.json files with patterns in the data folder.

The idea is to provide a simple list of files to be downloaded for applications that want a local cache of data definitions.

National Gallery Singapore https://www.nationalgallery.sg (((?:\d+)|([A-Z]+))-([A-Z0-9]+)(-[A-Z0-9.]+)*) 2003-03691 ASB-0036 2001-03415 RC-S2-CCS2.6 2010-04133

Musée d'Orsay https://www.musee-orsay.fr (([A-Z]+\ )(?:\d+)+((\ [0-9]+)|(,\ [A-z0-9\ ,]*)|(-[0-9]+))) RF 1949 17 RF 676, LUX 91 RF 1980-195 RF 37305, Recto, RF 37305

British Museum https://www.britishmuseum.org (([A-Z]+)([0-9]+)|([A-z]+.)(?:\d+)|(([A-z]?)(?:\d+),(?:\d+(.[0-9])([-0-9])*))) EA55614 1939,1010.1 2006,0503.1.1-20 EA24 Am.7184 Am1940,11.2

Create a tool that hashes the contents of all *.json files in the data directory and writes the value to a file.

The idea is to have a signal to applications can use to determine whether there are any changes to a local cache of data files that need to be downloaded.

Maybe install as a Git commit hook?

For example, the definition file for the Rijksmuseum assumes that the object_url property will point to the English-language version of their website:

    "object_url": "https://www.rijksmuseum.nl/en/collection/{accession_number}",    

Those kinds of assumptions shouldn't be... well, assumed. This might mean redefining the object_url property to be a dictionary along the lines of:

"object_url": {
   "en": "https://www.rijksmuseum.nl/en/collection/{accession_number}",
   "fr": "https://www.rijksmuseum.nl/fr/collection/{accession_number}"
}

Or maybe just changing the software packages that expand URI templates to take a dictionary of values (to expand) rather than a single (accession_number) value.

The latter approach feels more-better to me, as I write this. TBD...

This is a placeholder issue for definitions and tools to create "fake accession number" API endpoints for cultural heritage organizations that do not allow individual object pages to be accessed (directly) using an accession number. It's not clear whether that work should be part of this repo.

For example:

• https://www.nga.gov/collection/art-object-page.20483.html has accession number 1943.8.8285

And the data/objects.csv file in the NGA opendata release has the following header:

objectid,accessioned,accessionnum...

Which means a "fake accession number" API could map the accession number to the object ID and then either issue a redirect ot be used to explicitly assign a value to a object_uri URI template. For example:

• https://github.com/sfomuseum/accession-numbers/blob/main/schema/definition.schema.json#L28-L30

> git grep whosonfirst_id | grep -e '-1'
data/airandspace.si.edu.json:    "whosonfirst_id": -1,
data/americanhistory.si.edu.json:    "whosonfirst_id": -1,
data/artsmia.org.json:    "whosonfirst_id": -1,
data/chnmuseum.cn.json:    "whosonfirst_id": -1,
data/chrysler.org.json:    "whosonfirst_id": -1,
data/kanazawa21.jp.json:    "whosonfirst_id": -1,
data/mam.paris.fr.json:    "whosonfirst_id": -1,
data/musee-orsay.fr.json:    "whosonfirst_id": -1,
data/museivaticani.va.json:    "whosonfirst_id": -1,
data/museodelprado.es.json:    "whosonfirst_id": -1,
data/museoreinasofia.es.json:    "whosonfirst_id": -1,
data/museu.ms.json:    "whosonfirst_id": -1,
data/museum.go.kr.json:    "whosonfirst_id": -1,
data/nationalgallery.sg.json:    "whosonfirst_id": -1,
data/naturalhistory.si.edu.json:    "whosonfirst_id": -1,
data/nga.gov.json:    "whosonfirst_id": -1,
data/nmaahc.si.edu.json:    "whosonfirst_id": -1,
data/okeeffemuseum.org.json:    "whosonfirst_id": -1,
data/rusmuseum.ru.json:    "whosonfirst_id": -1,

Related: https://github.com/whosonfirst-data/whosonfirst-data/labels/glam