aboutsummaryrefslogtreecommitdiff
path: root/wqflask/base/data_set.py
diff options
context:
space:
mode:
Diffstat (limited to 'wqflask/base/data_set.py')
-rw-r--r--wqflask/base/data_set.py38
1 files changed, 15 insertions, 23 deletions
diff --git a/wqflask/base/data_set.py b/wqflask/base/data_set.py
index 2acb3b61..79f72390 100644
--- a/wqflask/base/data_set.py
+++ b/wqflask/base/data_set.py
@@ -171,28 +171,22 @@ class Markers(object):
def __init__(self, name):
json_data_fh = open(locate(name + ".json",'genotype/json'))
- try:
- markers = []
- with open(locate(name + "_snps.txt", 'r')) as bimbam_fh:
+ markers = []
+ with open("%s/%s_snps.txt" % (flat_files('genotype/bimbam'), name), 'r') as bimbam_fh:
+ if len(bimbam_fh.readline().split(", ")) > 2:
+ delimiter = ", "
+ elif len(bimbam_fh.readline().split(",")) > 2:
+ delimiter = ","
+ elif len(bimbam_fh.readline().split("\t")) > 2:
+ delimiter = "\t"
+ else:
+ delimiter = " "
+ for line in bimbam_fh:
marker = {}
- if len(bimbam_fh[0].split(", ")) > 2:
- delimiter = ", "
- elif len(bimbam_fh[0].split(",")) > 2:
- delimiter = ","
- elif len(bimbam_fh[0].split("\t")) > 2:
- delimiter = "\t"
- else:
- delimiter = " "
- for line in bimbam_fh:
- marker['name'] = line.split(delimiter)[0]
- marker['Mb']
- marker['chr'] = line.split(delimiter)[2]
- marker['cM']
- markers.append(marker)
- #try:
- # markers = json.load(json_data_fh)
- except:
- markers = []
+ marker['name'] = line.split(delimiter)[0].rstrip()
+ marker['Mb'] = float(line.split(delimiter)[1].rstrip())/1000000
+ marker['chr'] = line.split(delimiter)[2].rstrip()
+ markers.append(marker)
for marker in markers:
if (marker['chr'] != "X") and (marker['chr'] != "Y"):
@@ -334,8 +328,6 @@ class DatasetGroup(object):
return mapping_id, mapping_names
def get_markers(self):
- logger.debug("self.species is:", self.species)
-
def check_plink_gemma():
if flat_file_exists("mapping"):
MAPPING_PATH = flat_files("mapping")+"/"