MAX_POS too small

[SergejN]

Dear pysam authors,

I am trying to work with a large genome, but unfortunately, I am not able to convert my BAM files to bigwig using deeptools. Digging in the code lead me to the suspicion that the problem comes from the module pysam/libchtslib.pyx. Indeed, if you check the source code you see that MAX_POS is defined as 2^29 instead of 2^31 (BAM format): Line 44: cdef int MAX_POS = 2 << 29

Is it something that can be easily fixed?
I would appreciate that. Thank you!
Sergej

[AndreasHeger]

Certainly, this should be an easy fix.

[rembart]

hey i just had a look at the fix, and without being experienced in cython it seems to me that you just oversized the int32 - I would suggest to use an unsigned int or a long for the loops
otherwise the max_pos will default to "0" since 2^32 is not a valid "positive" int32 value

i haven't tested this but please have a look

best klaus

@AndreasHeger

[AndreasHeger]

Thanks, indeed, will need to do this more carefully on a separate branch.

[jmarshall]

The C code effectively uses (2 << 31) - 1 for this, which fits in an int32 and has the desired behaviour (modulo thinking about a couple of <= vs. < in the pysam code…).

[AndreasHeger]

Thanks, will do the same.

[jmarshall]

Uh, when I said the C code effectively uses (1 << 31) - 1, I meant it writes it as INT_MAX or 2147483647 or 0x7FFFFFFF or similar. Writing (1 << 31) - 1 overflows in C and perhaps should make you nervous in a .pyx file…

Perhaps this would be better using from libc.stdint cimport INT32_MAX as already is imported in several of these files?