search

GMOD / jbrowse

JBrowse 1, a full-featured genome browser built with JavaScript and HTML5. For JBrowse 2, see https://github.com/GMOD/jbrowse-components.
http://jbrowse.org
Other
464 stars 199 forks source link

Autocomplete Broken: "Error reading from name store." #513

Closed laceysanderson closed 10 years ago

laceysanderson commented 10 years ago

I've just created a JBrowse (v1.11.5) using Medicago truncatula data. I was able to load in the reference sequence, GFF3 gene data and even VCF (Medicago HapMAP) data successfully. However, once I ran the bin/generate-names.pl script I now get an alert box saying "Error reading from name store." Once I dismiss the box I'm still able to use the browser however the autocomplete of names does not work...

The bin/generate-names.pl did not output anything to the screen... When I look in my JBrowse installation, it did create a names directory in my data directory with a large number of subdirectories. One thing to note is that there is no names/root.json created even though I noticed the JBrowse/Browser.js script requires such a file.

Any ideas what might be causing this problem? ~Lacey

JBrowse exhibiting the problem: http://knowpulse2.usask.ca/jbrowse/JBrowse-1.11.5/?loc=chr1%3A23626611..23775389&tracks=DNA%2CGenes&highlight= OS: Debian 7 Apache 2.2.22

cmdcolin commented 10 years ago

From what I can see, the error will appear anytime there is a search box query for a chromosome name and position. So if I search "chr1:1..10000", then it will produce an error, but just "chr1" will not. This is pretty strange and I'm really not sure what causes that. For example, this isn't a problem on our test browser at http://jbrowse.org/code/JBrowse-1.11.5/?data=sample_data%2Fjson%2Fvolvox

So I guess I have to speculate now....

One thing I noticed on your browser is the number of hash bits currently use for the indexing is pretty high. It's sort of an obscure feature, but you could try specifying the number of hash bits explicitly with generate-names.pl --hashBits 16. Currently, your name store is using 28 hash bits which might actually be causing problems. JBrowse tries to estimate the right amount of bits but I think it may have gone overboard here, and you might want to force it to something around 12-16 which are more like default values. Just for reference, each 4 bits is one hex letter in the filename!

PS I am not sure if root.json is used anymore, but I didn't see any HTTP requests for this file so I don't think it is causing the error. There are however other 404 errors to the names store files but they have the hashed filenames, so that's what i'm focusing on here :)

laceysanderson commented 10 years ago

Good speculation! Re-running generate-names.pl with --hashBits 16 fixed it :)

cmdcolin commented 10 years ago

@laceysanderson I tried visiting earlier and it seemed to be working, but then I visited again and now it seems to give the error. Maybe it was not fixed?

laceysanderson commented 10 years ago

Hmm… I just checked again and it seems fine. I even checked on a bunch of difference references. Where were you in the genome when you got the error and was it the same error or different?

~Lacey

Lacey-Anne Sanderson Bioinformaticist Pulse Crop Breeding and Genetics Phone: (306) 966-3208 Email: laceyanne_sanderson@shaw.ca Room 2C33 Agriculture Department of Plant Sciences University of Saskatchewan

On Sep 18, 2014, at 12:03 PM, Colin notifications@github.com wrote:

@laceysanderson I tried visiting earlier and it seemed to be working, but then I visited again and now it seems to give the error. Maybe it was not fixed?

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

cmdcolin commented 10 years ago

I was on chr1. However, it may have been a crazy cache issue, because it is fixed again for me now.

cmdcolin commented 10 years ago

This is a tangent but I also added a small patch from a bug I noticed on your instance where VCF samples with "no data" called were displaying genotype wrong. I didn't know about this "no data" option so it was parsing genotype wrong.

Here's the patch https://github.com/GMOD/jbrowse/commit/db5c5fc6552881f52af0c29a1339cc7f65425539

laceysanderson commented 10 years ago

Oh, that's extra awesome :) Thanks! ~Lacey

Lacey-Anne Sanderson Bioinformaticist Pulse Crop Breeding and Genetics Phone: (306) 966-3208 Email: laceyanne_sanderson@shaw.ca Room 2C33 Agriculture Department of Plant Sciences University of Saskatchewan

On Sep 18, 2014, at 1:50 PM, Colin notifications@github.com wrote:

This is a tangent but I also added a small patch from a bug I noticed on your instance where VCF samples with "no data" called were displaying genotype wrong. I didn't know about this "no data" option so it was parsing genotype wrong.

Here's the patch db5c5fc

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