Closed andrewparkermorgan closed 6 years ago
hi Andrew,
That's so interesting. I just today got an email from another tximport user with another implementation of HDF5 imports.
This is not the active branch of tximport (which is the Bioconductor devel branch), but I'll work on porting this code over to the Bioconductor branch.
(Unfortunately, Bioconductor is not yet fully integrated with git, and uses svn as the primary version control.)
@andrewparkermorgan @DarwinAwardWinner you both sent me some HDF5 supporting code this week, and i'm working on incorporating it into the Bioconductor devel branch. right now i'm adding h5 files to the tximportData so that any new code will be covered by unit tests
The HDF5 file produced by
kallisto
is much smaller than the plain-text file even in the absence of bootstraps. For projects with many samples it would be advantageous to keep just the HDF5 file around. To that end I added a function that uses the Bioconductor packagerhdf5
to readkallisto
results from the HDF5 file. There is a modest speedup intximport()
and this could smooth the way to supporting bootstraps in the future if that is of interest.To minimize the number of dependencies for
tximport
, therhdf5
package is only added inSuggests
, and import dies without fallback if the package is not present.