People find the cactus header requirements + error messages confusing

[joelarmstrong]

Low priority, but: The requirements that cactus has for fasta headers are somewhat confusing to users, and it's especially bad since a lot of NCBI/genbank identifiers break the requirements. We should look into making the error messages more straightforward and provide a suggested solution (replacing .'s and/or spaces with underscores, probably).

We can't do much about the browser's requirements for the "." syntax, but we could look into using [nameparse=full] instead of [nameparse=darkspace] for lastz to attempt to get rid of the "first word of a header must be unique" rule. This would make parsing the cigars difficult, though, so it may be more trouble than it's worth.

[diekhans]

We should investigate why tge ucsc browser places these restrictions on sequence names. It might be unnecessary an loses important information

-----Original Message-----
From: Joel Armstrong notifications@github.com
To: benedictpaten/cactus cactus@noreply.github.com
Sent: Wed, 01 Oct 2014 9:53 AM
Subject: [cactus] People find the cactus header requirements + error messages confusing (#8)

Low priority, but: The requirements that cactus has for fasta headers are somewhat confusing to users, and it's especially bad since a lot of NCBI/genbank identifiers break the requirements. We should look into making the error messages more straightforward and provide a suggested solution (replacing .'s and/or spaces with underscores, probably).

We can't do much about the browser's requirements for the "." syntax, but we could look into using [nameparse=full] instead of [nameparse=darkspace] for lastz to attempt to get rid of the "first word of a header must be unique" rule. This would make parsing the cigars difficult, though, so it may be more trouble than it's worth.

---

Reply to this email directly or view it on GitHub:
#8

[joelarmstrong]

Brian said a while back they might cause problems because mySQL uses "." as a separator between db and table. Not sure if any of the assembly hub stuff actually goes through mySQL, but maybe it goes through code that is expecting a ".table" kind of syntax.

[diekhans]

They quit making tables spit by chromosome years ago. We should be able to
drop this restriction. It's especially bad for assemblyHubs where there
are no mysql tables.

Joel Armstrong notifications@github.com writes:

Brian said a while back they might cause problems because mySQL uses "." as a
separator between db and table. Not sure if any of the assembly hub stuff
actually goes through mySQL, but maybe it goes through code that is expecting a
".table" kind of syntax.

—
Reply to this email directly or view it on GitHub.*

[joelarmstrong]

OK, I've tested it a bit and it looks like there is no reason for that particular "ucscNames" restriction anymore. (Other than exported MAFs being hard to parse, but that isn't so bad.) Removing that will make things much better, there are so many headers with "."s in them that broke things before...

[joelarmstrong]

OK, we now ignore the "ucscNames" restriction by default, but enforce some requirements for characters that the UCSC browser still doesn't like, like "|". We should look into fixing the browser to allow "|" characters especially, since genbank sequences often have those in their identifiers.

The "first word must be unique" requirement is harder to get rid of, and it follows the FASTA spec, so I think it's more understandable.