increase buffer limit of dataread

I am trying to use the function affygcrma() in the bioinformatics toolbox. I have the proper input arguments and followed the error in the debugger to find the culprit function.

affygcrma calls affyprobeseqread

affyprobeseqread calls fastaread, makes seq_struct

affyprobeseqread calls handlefastatext(seq_struct)

handlefastatext makes large cell arrays: probe_headers, probe_seqs, probe_ids, probe_xs

for my sequence of interest, the cell array probe_headers size is 1x39070 handlefastatext loops through cell array index and calls strread(i)

inside strread, numel(probe_header) exceeds buffer limit of dataread(), which is 4095

function crashes:

*Error using dataread
Trouble reading integer from file (row 1, field 4) ==> NM_005505.2 /DB_XREF=gi:21361199 /GEN=SCARB1

Error in strread (line 51)
    [varargout{1:nlhs}]=dataread('string',varargin{:});

Error in handlefastatext (line 13)
    [chip_type, probe_ids(i), probe_xs(i)]= strread(probe_headers{i},...*

It looks like I cannot proceed unless I can change the buffer limit to 39071

Please help!!

Answer 1

[Walter Roberson]

dataread() is an undocumented mex file.

Although the comment in strread() implies that the 'bufsize' parameter to dataread is limited to 4095, my experiments do not seem to find that limit being enforced. Perhaps though it is the limit on the number of output values; I did not test that.

Anyhow, you could experiment by copying strread and editing the limit.

[Jan]

The source file dataread.c has been included in older Matlab versions. If you have access to one these versions, you can modify the source accordingly and recompile the function under a new name.

Answer 2

[Rory]

thanks walter

i bet you are right. it turned out that my bug was not actually due to this function after all.

[Walter Roberson]

Ah? What was the cause?