On Wed, Aug 13, 2014 at 7:22 PM, Eric Rasche <rasche.eric(a)yandex.ru> wrote:
I'm working on composite datatypes now (for PacBio SMRT cells). In the
datatype I know I'll have files with variable names (e.g. .1, .2, .3) and
after using the blast datattype as reference material, I noticed that you
had commented out files with variable names from the datatype. Is there
rationale behind that? Are we ensured that the entire directory will always
be transferred so it's not a problem to not specify files as part of the
Edward Kirton wrote the original BLAST DB datatypes - I'm not sure
how to nicely define open ended file lists for datatypes, but also the
makeblastdb wrapper output does not have this problem (yet).
If and when we try to support partitioned BLAST databases (with a
*.nal or *.pal alias file) then this would be needed.
I would think that the HTML composite datatype might be a better
guide here, since you will order get lots of child files (images etc).
My guess is the whole directory may be transferred anyway, but
having "undefined" files could be a problem at some point...