#! /bin/bash
# Last edited on 2008-06-13 18:10:46 by stolfi

# Usage: get-region-lengths {SETNAME} {SETDIR} {BASDIR} {OUTDIR}
# Extracts region length statistics for a set of labeled genes. 

# Reads from the file "{SETDIR}/{SETNAME}.gset" a list of full item
# names, of the form "{SPECIES}/{ITEM}" where {SPECIES} is the name of
# a species's dataser in the EID database (e.g. "at2004", "hs35p1")
# and {ITEM} is an EID item identifier (e.g. "10008_NC_003074").

# Then reads the label file "{BASDIR}/{SPECIES}/{ITEM}.lab" for all
# those items, and tabulates the lengths of exon and non-exon regions
# in them. Writes the statistical parameters (average length, etc.)
# and the length histogram to the file
# "{OUTDIR}/{SETNAME}-{TYPE}.lens" where {TYPE} is "K" or "N".

setName="$1"; shift
setDir="$1"; shift
basDir="$1"; shift
outDir="$1"; shift

setFile="${setDir}/${setName}.gset"
outName="${outDir}/${setName}"

dataBank="/tmp/$$"
progdir=${STOLFIHOME}/programs/c/DNA/dnabayes

cat ${setFile} \
  | gawk \
      -v basDir="${basDir}" \
      '/^[a-zA-Z0-9]/{ printf "%s/%s.lab\n", basDir, $1; }' \
  > ${dataBank}

${progdir}/dbd_gather_region_length_stats \
    -v maxLength=2999 \
    -v outName=${outName} \
    `cat ${dataBank}` 

/bin/rm -f ${dataBank}