I really need to update my local copy. I must admit, I'm a bit afraid to do so.

On 4/7/11 11:35 AM, Kelly Vincent wrote:
That opening-the-entire-file inefficiency in our sam_to_bam.py was fixed
quite some time ago (4/30/2010, in 3724:007f175c8b88). It has used
getsize since then. It's really just intended to catch weird errors that
don't throw an actual error (also, I think that if the sam file has no
header, no info would be output into the bam file). It seems somewhat
more informative to tell the user the file is empty rather than just
"successfully" output an empty file.

Kelly


On Apr 7, 2011, at 1:05 AM, Assaf Gordon wrote:

Just another example why python's misleadingly simple idioms are quite
dangerous in production code (couldn't help myself from teasing about
python... sorry about that).

Seems like line 150 in "sam_to_bam.py" tries to read the entire BAM
file into memory just to find out if it's empty or not...

As a stop gap solution with minimal changes, change line 150 from:
if len( open( tmp_aligns_file_name ).read() ) == 0:
to
if len( open( tmp_aligns_file_name ).read(10) ) == 0:

Which will read up to the first 10 bytes (instead of the entire file).

A slightly better (but still wrong) solution is to simply check the
file size, with:
if os.path.getsize(tmp_aligns_file_name) == 0:

But it's still wrong because even an invalid sam file will create a
non-empty BAM file (when using "samtools view -bt") - the BAM file
will still contain the chromosome names and sizes.

Example:
========
$ cat mm9.fa.fai
chr1 197195432 6 50 51
chr10 129993255 201139354 50 51
chr11 121843856 333732482 50 51
chr12 121257530 458013223 50 51
chr13 120284312 581695911 50 51
chr13_random 400311 704385924 50 51
chr14 125194864 704794249 50 51
chr15 103494974 832493018 50 51
...
...

$ cat 1.sam
Hello World
This is not a SAM file

$ samtools view -bt mm9.fa.fai -o 1.bam 1.sam
[sam_header_read2] 35 sequences loaded.
[sam_read1] reference 'This is not a SAM file' is recognized as '*'.
[main_samview] truncated file.

$ ls -l 1.*
-rw-r--r-- 1 gordon hannon 348 Apr 7 00:57 1.bam
-rw-r--r-- 1 gordon hannon 35 Apr 7 00:57 1.sam
========


So in short, this whole sam-to-bam wrapper tool is not suitable for
large SAM files (if they don't fit entirely in memory), and not for
error checking of invalid SAM files.


-gordon


On 04/07/2011 12:30 AM, Ryan Golhar wrote:
Here's what I get:

(galaxy_env)[galaxy@vail pbs]$ sh ./big.sh
Samtools Version: 0.1.14 (r933:170)
Traceback (most recent call last):
File "/home/galaxy/galaxy-dist/tools/samtools/sam_to_bam.py", line 150,
in __main__
if len( open( tmp_aligns_file_name ).read() ) == 0:
MemoryError
Error extracting alignments from
(/home/galaxy/galaxy-dist/database/files/000/dataset_785.dat),
(galaxy_env)[galaxy@vail pbs]$



On 4/6/11 7:29 PM, Assaf Gordon wrote:
Ryan,

Since we're shooting in the dark here, best to try and understand
what's the exception.

Add the following line to the beginning of "sam_to_bam.py":
import traceback

and add the following line to "sam_to_bam.py" line 156 (before the
call to "stop_err"):
traceback.print_exc()

Hopefully this will print out which exception you're getting, and
where is it thrown from.

-gordon

___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client. To manage your subscriptions to this
and other Galaxy lists, please use the interface at:

http://lists.bx.psu.edu/



--
CONFIDENTIALITY NOTICE: This email communication may contain private, confidential, or legally privileged information intended for the sole use of the designated and/or duly authorized recipient(s). If you are not the intended recipient or have received this email in error, please notify the sender immediately by email and permanently delete all copies of this email including all attachments without reading them. If you are the intended recipient, secure the contents in a manner that conforms to all applicable state and/or federal requirements related to privacy and confidentiality of such information.

<<attachment: golharam.vcf>>

___________________________________________________________
Please keep all replies on the list by using "reply all"
in your mail client.  To manage your subscriptions to this
and other Galaxy lists, please use the interface at:

  http://lists.bx.psu.edu/

Reply via email to