Recursive dependency

[chorny]

Recursive dependency detected:
    Bio::Variation::SNP (have: N/A; want: 1.007004)
 => CJFIELDS/Bio-Variation-1.7.4.tar.gz
 => Bio::DB::GenBank (have: N/A; requires: 0)
 => CJFIELDS/Bio-DB-NCBIHelper-1.7.6.tar.gz
 => Bio::SeqIO::entrezgene (have: N/A; requires: 0)
 => CJFIELDS/Bio-ASN1-EntrezGene-1.73.tar.gz
 => Bio::Cluster::SequenceFamily (have: N/A; requires: 0)
 => CJFIELDS/Bio-Cluster-1.7.3.tar.gz
 => Bio::Variation::SNP (have: N/A; requires: 0).

[cjfields]

See also bioperl/bioperl-live#331

[cjfields]

We'll have to trace through this and see specifically where we can break this dependency. If it's only for tests this could be set up for running in dev-only mode.

[carandraug]

The dependency on Bio::DB::GenBank and Bio::DB::EMBL is for the bp_flanks program so it can be called with an accession number instead of a file.

[cjfields]

@carandraug ah, makes sense. Question is how to disentangle this; any recommendations?

[carandraug]

One suggestion is to make the class to use an argument to bp_flanks. This makes it a separate dependency, it makes a dependency that the user must explicity name, and also enables bp_flanks to work with any class that implements a specific interface.

[cjfields]

@carandraug seems like a reasonable solution. I did notice that the script doesn't actually use Bio::Variation modules, it is primarily used for extracting flanking regions for any list of positions (the most obvious case being variants) from local or remote sequence records. Any reason it's in this dist beyond it involving variants?

[carandraug]

The commit messages bioperl/bioperl-live@1f7b54a says:

Bio::Variation::* move namespace into its own distribution.

Also move the program bin_flanks which although not actually using the
Bio::Variation modules, it's to be used in the same context.

[cjfields]

@carandraug works for me (makes sense to do so, just wanted to make sure).

[cjfields]

Pushing out a new point release (1.7.5), will see if this addresses the problem.

[carandraug]

The change in d3a1dcc is really not what I suggested. But if you want to go that way, you might was well not change the code and list those two modules as dependencies to ignore on dist.ini with AutoPrereqs.skips (see bioperl-live dist.ini for an example)

[carandraug]

[...] you might was well not change the code [...]

Well, you would only need to change the use to a require within the right block.

[cjfields]

Well, you would only need to change the use to a require within the right block.

The fix was a bit low-hanging fruit but using the BioPerl API, we could switch it to a simple eval "require $class" or similar.

There will be instances where installing this distribution directly will fail with the auto-generated t/00-compile.t tests, see: http://www.cpantesters.org/cpan/report/1dcbc584-209d-11ea-99fe-b031dbec7dbf. Not sure if ignoring within dist.ini fixes this?