Better handle shelfmarks for closed stores items

[jtweed]

Many of the closed stores items in the catalogue have a shelfmark. Some of these are really reference numbers and some are store locations that we don't want to display publicly as shelfmarks on the site.

There will be examples of each of the following:

• Shelfmarks that we want to suppress completely from public view
• Shelfmarks that we want to display as the referenceNumber on the work (and also add as an identifier)
• Shelfmarks that we want to make a kind of identifier on the work
• Shelfmarks that we want to make a kind of identifier on the item

By ensuring that these shelfmarks that are really identifiers are properly added as such, we can:

• Display the reference numbers that we encourage researchers to use for certain items eg iconographic numbers
• Ensure that they are picked up in search, so that they can be used to find a particular work
• Use them to look up linking entries to related works that we hold in the collection

First step here is analysis of the existing closed stores shelfmarks in collaboration with Collections Information. It's likely that the resulting rules may be based on a combination of material types, location codes and pattern matching.

---

• [x] Trim whitespace from shelfmark values
• [x] Start pulling the shelfmark from 949 $a, not the callNumber field

[alexwlchan]

Link to a corresponding Slack thread: https://wellcome.slack.com/archives/CUA669WHH/p1619194557197800

[alexwlchan]

There's a script in the API repo that lets you build your own spreadsheet of shelfmarks: https://github.com/wellcomecollection/catalogue-api/blob/main/snapshots/analysis/get_shelfmark_tally.py

[alexwlchan]

I started looking through this, and in particular looking for the longest examples of shelfmarks (which has implications for the UI):

['750339i, 750340i, 750341i, 750374i, 750376i',
 '751207i, 751208i, 751209i, 751210i, 751211i',
 '751314i, 751315i, 751316i, 751317i, 751318i',
 '751407i, 751408i, 751410i, 751411i, 751412i',
 '750501i, 750518i, 750519i, 750520i, 750521i',
 '751232i, 751237i, 751238i, 751239i, 751240i',
 'Entrance Hall, Wellcome Library, 2015 - 2020',
 'Bound with Underwood, M. Surgical tracts. 1788',
 'New Research Institute, University of Edinburgh',
 '/BIO [spine reads Obituary notices ... 1898-1904]',
 '14084i (Browning) - 14085i (Herschel) on one mount',
 '750766i, 750767i, 750769i, 750771i, 750773i, 750774i',
 '750642i, 750643i, 750644i, 750645i, 750646i, 750648i',
 '750489i, 750490i, 750491i, 750492i, 750493i, 750494i',
 '751374i, 751375i, 751376i, 751377i, 751378i, 751379i',
 '751380i, 751381i, 751382i, 751383i, 751384i, 751385i',
 '751163i, 751164i, 751165i, 751166i, 751167i, 751168i',
 '750760i, 750761i, 750762i, 750763i, 750764i, 750765i',
 '751442i, 751457i, 751458i, 751459i, 751460i, 751461i',
 '750623i, 750624i, 750625i, 750626i, 750627i, 750628i',
 '751429i, 751430i, 751431i, 751432i, 751433i, 751439i',
 '751201i, 751202i, 751203i, 751204i, 751205i, 751206i',
 '750754i, 750755i, 750756i, 750757i, 750758i, 750759i',
 '751299i, 751300i, 751301i, 751302i, 751303i, 751304i',
 '751266i, 751267i, 751268i, 751269i, 751270i, 751271i',
 '750377i, 750378i, 750379i, 750380i, 750381i, 750382i',
 'Entrance Hall, Wellcome Trust, 215 Euston Road, London',
 'Medicine Man, Wellcome Collection, 9 March - July 2020',
 'Medicine Man, Wellcome Collection, February 2015 onwards',
 '2897011i and 2897018i in one bifolium attached to 2889844i',
 '2535399i, 2535400i, 2535401i, 2535402i, 2535403i, 2535404i,',
 'Being Human, Wellcome Collection, 5 September 2019 - ongoing',
 'WF/E/11/01 (part 1):Box AMS/MF/171 (original item in box 23)',
 'WF/E/11/01 (part 2):Box AMS/MF/172 (original item in box 23)',
 'WF/E/11/01 (part 3):Box AMS/MF/173 (original item in box 23)',
 '751414i, 751415i, 751416i, 751417i, 751418i, 751419i, 751420i',
 'Claret Wall Display, Wellcome Library, 12 May 2014 - 12 May 2019',
 'Medicine Galleries, Science Museum, London, November 2019 - ongoing',
 'Entrance Hall, Wellcome Library, 20 February 2018 - 20 February 2019',
 '751347i, 751348i, 751349i, 751350i, 751351i, 751352i, 751353i, 751354i',
 'Bodyscapes, The Israel Museum, Jerusalem, 21 February - 3 October 2020',
 '209812i volume of documents: Dental memoranda compilied by Theodosius Purland',
 '751367i, 751368i (two impressions), 751369i, 751370i, 751371i, 751372i, 751373i',
 '/LEATHER                                                                       ',
 'George III: The Mind Behind the Myth, Kew Palace, London, 3 April - 27 September 2020',
 "Unearthed: Photography's Roots, Dulwich Picture Gallery, 2 December 2020 - 9 May 2021  ",
 '751272i, 751273i, 751274i, 751269i, 751270i, 751271i, 751295i, 751296i, 751297i, 751298i',
 'Tantra: Enlightenment to Revolution, British Museum, London, 24 September 2020 - 24 January 2021',
 'Masculinities: Liberation through Photography, Barbican Art Gallery, London, 20 February - 17 May 2020',
 'Passions and Fervour.  The Art of Powerful Emotions, LWL- Museum fur Kunst und Kultur, Munster, Germany,  9 October 2020 - 14 F']

Two immediate observations:

• We should trim whitespace from shelfmarks (e.g. '/LEATHER ')
• There are 15,501 unique shelf marks (on 18,845 items) that are just a comma-separated list of identifiers like 751414i. I know this is some form of identifier (although I don't know what it's called). We could pull this out as a separate list. I'm going to have a look at where this value is pulled from.

[alexwlchan]

Following up from a discussion in Slack:

We're going to swap out the callNumber field for 949 $a, and trim whitespace while we're here.

[jtweed]

I think the ones that look like where something is on exhibition should possibly be a location label rather than a shelfmark?

[alexwlchan]

Yeah, we can look at pulling 999 into the label field.

[alexwlchan]

We have ~16k unique "EPB*" identifiers in the shelfmark field, that feel like a plausible candidate for handling similar to i numbers.

We should talk to Collections to find out what else is a wolf-in-shelfmark's clothing.