Better documentation of, &/or support for, other platforms

[mjwestgate]

As of v1.4.0, galah supports the ALA (default), and 5 other atlases (Austria, Guatemala, Spain, Sweden & UK). Although other atlases exist (see the full list), it is unclear whether these other atlases are unsupported because of technical issues, or because we simply haven't added them yet.

In increasing order of difficulty, a useful set of activities to address this might include:

1. [SIMPLE] List all living atlases within show_all_atlases(), and record whether they can be called by galah
2. Allow other atlases to be queried by data-retrieving functions such as atlas_counts and atlas_occurrences
3. Support queries to taxonomic name services of non-ALA atlases via search_taxa or search_identifiers, rather than requiring the user to search using taxize. This could work either by importing taxize, or by writing new code
4. [DIFFICULT] Support calls to the GBIF API using galah functions. This would require galah_filter to support queries other than solr. We would also need to either call rgbif::occ within atlas_occurrences to download the data, or write new code to achieve the same goal.

[mjwestgate]

As of commit a0e0b2e504b6e21271018a555aee85a2ddb974c4, we now support name matching for all atlases that use GBIF taxonomy (currently Austria, Guatemala, Spain & Sweden). This effectively means we can remove the advise to use taxize for this purpose from our documentation. What we can't do yet is query other GBIF APIs from galah, which leaves open the question of what to do next. Basically we have two options:

• Only support the living atlas community, by supporting name matching services from GBIF, but no other GBIF data/services
• Expand galah to retrieve record counts or occurrences from GBIF

Based on my reading of the documentation here, there are a few points to consider in reaching a decision. For occurrence data:

• There is no API for field values, meaning that show_all_values/search_values won't work unless we hard-code some entries. This is effectively the approach taken by e.g. rgbif::isocodes
• GBIF don't allow the user restrict the columns returned in an occurrence download, meaning that galah_select would be either be ignored or applied post-download (which is inefficient)
• The occurrence download process for GBIF is fairly involved, but does support filtering on a large number of fields. The syntax is different from solr, in that it requires a JSON-type structure which we don't currently produce; but it could be made to work
• Downloads appear limited to 101k records, much less than the 50 million allowed by ALA. This is somewhat irrelevant as clearly people still want information from GBIF; but would require we update error messages etc.

For counts the same issues apply, plus a few more:

• The GBIF counts API only supports a narrow range of facets (effectively only year and basisOfRecord), which limits usefulness of galah_group_by, unless we write our own iteration code
• There is a hack to apply occurrence filters on counts, by using the occurrence/search API and adding &limit=0. This would maintain equivalence between atlas_counts and atlas_occurrences, which is an important feature.

In summary, while there is merit in supporting count and occurrence queries from GBIF; but the architecture is sufficiently different to make this somewhat challenging.

[mjwestgate]

There is a new Living Atlases website that lists current participants and their status here. GBIF 'noted portals' are available here, but it is unclear if there are other sites that use GBIF infrastructure hosted elsewhere.

[mjwestgate]

This is closed, with the caveat that different atlases perform differently, and GBIF not currently supported