fsprojects / bioproviders Goto Github PK

At present, the type provider relies on a list of FTP locations for the different species and assemblies on GenBank. These will become outdated over time as GenBank updates, so it would be useful to include a script in the repository that allows us to update the lists every so often (and then publish a new version of the package).

Alex has noted that the files generated are based off a file on the GenBank FTP server itself; in future we may be able to look into using this directly rather than keeping our own lists. This could help us avoid the problems in #6.

Are the docs published to gh-pages?

An example where we do this auto-magically on every push to main is here: https://github.com/fsprojects/FSharp.Data/blob/main/.github/workflows/push-master.yml#L25-L31

Consider auto-publishing a nuget off pushes to main?

For example like here: https://github.com/fsprojects/FSharp.Data/blob/main/.github/workflows/push-master.yml#L34-L35

Previously, when building the package for NuGet, the files with the lists of locations on the GenBank FTP server would be placed in the same directory as the DLLs, rather than in the data subfolder. This meant that they could not be found when using the type provider from NuGet.

At present, I have moved the files out of the subfolder to attempt to avoid this problem (though my attempt to use the package locally still did not work, for other reasons). There are a few different ways we could tackle this problem:

• Find a way to properly put the files in the correct subfolder while letting the type provider still work as a project reference, if that's still desired;
• Make the data files and access code a separate package, as the files can be more easily put into a subfolder this way; or
• Remove the dependency on these data files entirely, as mentioned in #5.

Looks good - but time to migrate the docs to the main project README.

There are a couple of issues with the current generation of documentation using fsdocs. I've spent some time trying to figure these out, but have not had any success yet.

• When running the build script, the code in GenBankExample.fsx does not run. The output of the build script mentions that the DesignTime assembly is already loaded and without being able to load the BioProviders package, the code naturally doesn't work.
• I cannot get the License, Release Notes and Source Repository links to have the correct URL. I've tried including a Directory.Build.props like the original FSharp.Formatting in the solution (which isn't committed yet) with the required XML properties but that doesn't seem to work.

At present, the type provider only supports GenBank, though there are routes to use RefSeq data that lead to "unsupported" message outputs. Once we are happy with the state of the GenBank type provider, we may look into creating a RefSeq type provider too.

At present, the BioProviders examples only show how to use the type provider in an .fsx script file. Since we are publishing on NuGet, it will likely be used in Visual Studio projects too, and we should include examples of doing that.

According to Alex, the Metadata in the GenBankFlatFile type does not provide all the available metadata fields. We need to check whether there are any useful ones (such as Locus) that are not included yet and add them to the type.

I'm wondering if https://github.com/fslaborg might be a better home for this? Or https://github.com/fsprojects?