Closed xflr6 closed 1 year ago
The core problem seems to be incomplete bibliographical metadata, which tricks the merging algo into thinking a lot of refs are the same, e.g.
@book{silpng1991-2000:462,
author = {___ ___ and trans. ___ and trans.},
title = {I-unöý öýgöbin [preschool primer 3,4]},
pages = {42},
year = {1994},
glottolog_ref_id = {158690},
lgcode = {Bunama [bdd]},
macro_area = {Papua}
}
@book{silpng1991-2000:463,
author = {___ ___ and trans. ___ and trans.},
title = {I-unöý öýgöbin [preschool primer 5]},
pages = {32},
year = {1994},
glottolog_ref_id = {158690},
lgcode = {Bunama [bdd]},
macro_area = {Papua}
}
Now if we want to be able to resolve all local bibkeys (even multiple ones for the same provider and the same merged ref), we have to keep the unique constraint as loose as it is, it seems.
AFAICT, at least the web app does the right thing here: https://glottolog.org/resource/reference/id/158690
Not sure whether we intend to include the
id
column (which is always unique) within the uniqueness here: https://github.com/clld/glottolog3/blob/c0ed353dacfb62cd42bf779c0874878a023dfeed/glottolog3/models.py#L126-L130 Looks like this in the table (use theid
column as anotherscrtrickle
?):A small number of refs gets many provider links this way: