Skip to content

5.12.1

Choose a tag to compare

@aviezerl aviezerl released this 01 Oct 18:28
· 590 commits to master since this release
57f4181
  • Behavior fix: gcluster.run() jobs run against the caller's database, working directory, datasets and virtual tracks, and load misha from the caller's library. They had used the example database unless the job called gsetroot() itself.

  • Behavior fix: gtrack.liftover() could read an indexed source track from the wrong chromosomes, with no error: any 2D track, and 1D tracks in a per-chromosome database whose chrom_sizes.txt names lack the "chr" prefix. Re-lift such tracks. An indexed source track outside a database is now an error.

  • Behavior fix: gtrack.liftover() of a 2D points track (e.g. Hi-C) writes a points track instead of 1bp rectangles, with memory bounded per chromosome pair (gmax.data.size), so genome-scale Hi-C tracks can be lifted.

  • Behavior fix: gtrack.2d.import() and gtrack.2d.import_contacts() reject a negative start2. Such rows were silently dropped or moved.

  • gtrack.2d.import() and gtrack.2d.import_contacts() run in parallel (up to gmax.processes workers) with identical output. A chromosome pair starts only when its estimated memory fits within gmax.mem.usage; its default is 80% of the machine's RAM, so on a cluster set it to the job's share. gmultitasking = FALSE restores the serial import.

  • The 2D imports no longer crash R on an input file they cannot open or with gmax.data.size below 1.

  • An integer gmax.mem.usage above about 2 GB (e.g. 4000000L) overflowed; it is now read correctly.