needle -filter

Peter Rice pmr at ebi.ac.uk
Tue Jun 10 14:47:31 UTC 2003


Jan T. Kim wrote:
> More seriously: The standard input is logically one input stream. The
> issue here is not multiple input streams which, in the Unix environment,
> would typically be handled by using named pipes (this may also be an
> approach for solving the problem from which this thread originated).
> In many cases, though, it is preferable to have one input stream which
> is structured into multiple records, such as sequence entries in our
> case. Handling such multiple-record input would be more convenient if
> EMBOSS did not have to perform a pass through an entire stream for each
> record it needs to obtain. This approach is restricted to streams which
> can be "reproduced", which does not generally apply to the standard
> input stream.


EMBOSS does not (always) read the entire stream - but it would have to 
rewind the input to read a second sequence.

Consider, for example, FASTA format:

 >abc
aaacccgggttt
 >def
tttgggcccaaa

EMBOSS has to read the first line of the second sequence - just to know 
that the first sequence has ended.

Peter




More information about the EMBOSS mailing list