diff options
author | Arthur Centeno | 2021-06-15 15:33:59 +0000 |
---|---|---|
committer | Arthur Centeno | 2021-06-15 15:33:59 +0000 |
commit | c1a6ca69f7c48d99b6c5d62e56a445583fd4c08b (patch) | |
tree | f98fccc44829e3f52e585d6b41a1942aa52bd8b7 /scripts/maintenance | |
parent | 7e49c006af9c4f7453c3578a7d4f1fc4d7bdf3ed (diff) | |
parent | 9e9e0e4d440383f617542e810a1115833eafd7bf (diff) | |
download | genenetwork2-c1a6ca69f7c48d99b6c5d62e56a445583fd4c08b.tar.gz |
Merge branch 'testing' of github.com:genenetwork/genenetwork2 into acenteno
Diffstat (limited to 'scripts/maintenance')
-rwxr-xr-x | scripts/maintenance/datastructure.py | 177 | ||||
-rwxr-xr-x | scripts/maintenance/load_phenotypes.py | 43 | ||||
-rw-r--r-- | scripts/maintenance/utilities.py | 89 |
3 files changed, 291 insertions, 18 deletions
diff --git a/scripts/maintenance/datastructure.py b/scripts/maintenance/datastructure.py new file mode 100755 index 00000000..9f3e8b1e --- /dev/null +++ b/scripts/maintenance/datastructure.py @@ -0,0 +1,177 @@ +import utilities + +def get_probesetfreezes(inbredsetid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT ProbeSetFreeze.`Id`, ProbeSetFreeze.`Name`, ProbeSetFreeze.`FullName` + FROM ProbeSetFreeze, ProbeFreeze + WHERE ProbeSetFreeze.`ProbeFreezeId`=ProbeFreeze.`Id` + AND ProbeFreeze.`InbredSetId`=%s + """ + cursor.execute(sql, (inbredsetid)) + return cursor.fetchall() + +def get_probesetfreeze(probesetfreezeid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT ProbeSetFreeze.`Id`, ProbeSetFreeze.`Name`, ProbeSetFreeze.`FullName` + FROM ProbeSetFreeze + WHERE ProbeSetFreeze.`Id`=%s + """ + cursor.execute(sql, (probesetfreezeid)) + return cursor.fetchone() + +def get_strains(inbredsetid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT Strain.`Id`, Strain.`Name` + FROM StrainXRef, Strain + WHERE StrainXRef.`InbredSetId`=%s + AND StrainXRef.`StrainId`=Strain.`Id` + ORDER BY StrainXRef.`OrderId` + """ + cursor.execute(sql, (inbredsetid)) + return cursor.fetchall() + +def get_inbredset(probesetfreezeid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT InbredSet.`Id`, InbredSet.`Name`, InbredSet.`FullName` + FROM InbredSet, ProbeFreeze, ProbeSetFreeze + WHERE InbredSet.`Id`=ProbeFreeze.`InbredSetId` + AND ProbeFreeze.`Id`=ProbeSetFreeze.`ProbeFreezeId` + AND ProbeSetFreeze.`Id`=%s + """ + cursor.execute(sql, (probesetfreezeid)) + return cursor.fetchone() + +def get_species(inbredsetid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT Species.`Id`, Species.`Name`, Species.`MenuName`, Species.`FullName` + FROM InbredSet, Species + WHERE InbredSet.`Id`=%s + AND InbredSet.`SpeciesId`=Species.`Id` + """ + cursor.execute(sql, (inbredsetid)) + return cursor.fetchone() + +def get_genofreeze_byinbredsetid(inbredsetid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT GenoFreeze.`Id`, GenoFreeze.`Name`, GenoFreeze.`FullName`, GenoFreeze.`InbredSetId` + FROM GenoFreeze + WHERE GenoFreeze.`InbredSetId`=%s + """ + cursor.execute(sql, (inbredsetid)) + return cursor.fetchone() + +def get_nextdataid_genotype(): + cursor, con = utilities.get_cursor() + sql = """ + SELECT GenoData.`Id` + FROM GenoData + ORDER BY GenoData.`Id` DESC + LIMIT 1 + """ + cursor.execute(sql) + re = cursor.fetchone() + dataid = re[0] + dataid += 1 + return dataid + +def get_nextdataid_phenotype(): + cursor, con = utilities.get_cursor() + sql = """ + SELECT PublishData.`Id` + FROM PublishData + ORDER BY PublishData.`Id` DESC + LIMIT 1 + """ + cursor.execute(sql) + re = cursor.fetchone() + dataid = re[0] + dataid += 1 + return dataid + +def get_nextorderid_strainxref(inbredsetid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT StrainXRef.`OrderId` + FROM StrainXRef + WHERE StrainXRef.`InbredSetId`=%s + ORDER BY StrainXRef.`OrderId` DESC + LIMIT 1 + """ + cursor.execute(sql, (inbredsetid)) + re = cursor.fetchone() + if re: + orderid = re[0] + 1 + else: + orderid = 1 + return orderid + +def insert_strain(inbredsetid, strainname): + speciesid = get_species(inbredsetid)[0] + cursor, con = utilities.get_cursor() + sql = """ + INSERT INTO Strain + SET + Strain.`Name`=%s, + Strain.`Name2`=%s, + Strain.`SpeciesId`=%s + """ + cursor.execute(sql, (strainname, strainname, speciesid)) + +def insert_strainxref(inbredsetid, strainid): + orderid = get_nextorderid_strainxref(inbredsetid) + cursor, con = utilities.get_cursor() + sql = """ + INSERT INTO StrainXRef + SET + StrainXRef.`InbredSetId`=%s, + StrainXRef.`StrainId`=%s, + StrainXRef.`OrderId`=%s, + StrainXRef.`Used_for_mapping`=%s, + StrainXRef.`PedigreeStatus`=%s + """ + cursor.execute(sql, (inbredsetid, strainid, orderid, "N", None)) + +def get_strain(inbredsetid, strainname): + speciesid = get_species(inbredsetid)[0] + cursor, con = utilities.get_cursor() + sql = """ + SELECT Strain.`Id`, Strain.`Name` + FROM Strain + WHERE Strain.`SpeciesId`=%s + AND Strain.`Name` LIKE %s + """ + cursor.execute(sql, (speciesid, strainname)) + return cursor.fetchone() + +def get_strainxref(inbredsetid, strainid): + cursor, con = utilities.get_cursor() + sql = """ + SELECT StrainXRef.`StrainId` + FROM StrainXRef + WHERE StrainXRef.`InbredSetId`=%s + AND StrainXRef.`StrainId`=%s + """ + cursor.execute(sql, (inbredsetid, strainid)) + return cursor.fetchone() + +def get_strain_sure(inbredsetid, strainname, updatestrainxref=None): + strain = get_strain(inbredsetid, strainname) + if not strain: + insert_strain(inbredsetid, strainname) + strain = get_strain(inbredsetid, strainname) + strainxref = get_strainxref(inbredsetid, strain[0]) + if not strainxref and updatestrainxref: + insert_strainxref(inbredsetid, strain[0]) + return strain + +def get_strains_bynames(inbredsetid, strainnames, updatestrainxref=None): + strains = [] + for strainname in strainnames: + strains.append(get_strain_sure(inbredsetid, strainname, updatestrainxref)) + return strains diff --git a/scripts/maintenance/load_phenotypes.py b/scripts/maintenance/load_phenotypes.py index 759d2eec..aa02d0cd 100755 --- a/scripts/maintenance/load_phenotypes.py +++ b/scripts/maintenance/load_phenotypes.py @@ -1,3 +1,11 @@ +# Load Python3 environment with GN2 utilities: +# +# source /usr/local/guix-profiles/gn-latest-20210512/etc/profile +# +# and run +# +# python load_phenotypes.py [args...] + import sys import csv @@ -9,35 +17,34 @@ def main(argv): config = utilities.get_config(argv[1]) print("config:") for item in config.items('config'): - print(("\t%s" % (str(item)))) + print("\t%s" % (str(item))) # var inbredsetid = config.get('config', 'inbredsetid') - print(("inbredsetid: %s" % inbredsetid)) + print("inbredsetid: %s" % inbredsetid) species = datastructure.get_species(inbredsetid) speciesid = species[0] - print(("speciesid: %s" % speciesid)) + print("speciesid: %s" % speciesid) dataid = datastructure.get_nextdataid_phenotype() - print(("next data id: %s" % dataid)) + print("next data id: %s" % dataid) cursor, con = utilities.get_cursor() # datafile datafile = open(config.get('config', 'datafile'), 'r') phenotypedata = csv.reader(datafile, delimiter='\t', quotechar='"') - phenotypedata_head = next(phenotypedata) - print(("phenotypedata head:\n\t%s" % phenotypedata_head)) + phenotypedata_head = phenotypedata.next() + print("phenotypedata head:\n\t%s" % phenotypedata_head) strainnames = phenotypedata_head[1:] strains = datastructure.get_strains_bynames(inbredsetid=inbredsetid, strainnames=strainnames, updatestrainxref="yes") # metafile metafile = open(config.get('config', 'metafile'), 'r') phenotypemeta = csv.reader(metafile, delimiter='\t', quotechar='"') - phenotypemeta_head = next(phenotypemeta) - print(("phenotypemeta head:\n\t%s" % phenotypemeta_head)) - print() + phenotypemeta_head = phenotypemeta.next() + print("phenotypemeta head:\n\t%s" % phenotypemeta_head) # load for metarow in phenotypemeta: # - datarow_value = next(phenotypedata) - datarow_se = next(phenotypedata) - datarow_n = next(phenotypedata) + datarow_value = phenotypedata.next() + datarow_se = phenotypedata.next() + datarow_n = phenotypedata.next() # Phenotype sql = """ INSERT INTO Phenotype @@ -67,7 +74,7 @@ def main(argv): )) rowcount = cursor.rowcount phenotypeid = con.insert_id() - print(("INSERT INTO Phenotype: %d record: %d" % (rowcount, phenotypeid))) + print("INSERT INTO Phenotype: %d record: %d" % (rowcount, phenotypeid)) # Publication publicationid = None # reset pubmed_id = utilities.to_db_string(metarow[0], None) @@ -81,7 +88,7 @@ def main(argv): re = cursor.fetchone() if re: publicationid = re[0] - print(("get Publication record: %d" % publicationid)) + print("get Publication record: %d" % publicationid) if not publicationid: sql = """ INSERT INTO Publication @@ -109,7 +116,7 @@ def main(argv): )) rowcount = cursor.rowcount publicationid = con.insert_id() - print(("INSERT INTO Publication: %d record: %d" % (rowcount, publicationid))) + print("INSERT INTO Publication: %d record: %d" % (rowcount, publicationid)) # data for index, strain in enumerate(strains): # @@ -158,14 +165,14 @@ def main(argv): cursor.execute(sql, (inbredsetid, phenotypeid, publicationid, dataid, "")) rowcount = cursor.rowcount publishxrefid = con.insert_id() - print(("INSERT INTO PublishXRef: %d record: %d" % (rowcount, publishxrefid))) + print("INSERT INTO PublishXRef: %d record: %d" % (rowcount, publishxrefid)) # for loop next dataid += 1 - print() + print # release con.close() if __name__ == "__main__": - print(("command line arguments:\n\t%s" % sys.argv)) + print("command line arguments:\n\t%s" % sys.argv) main(sys.argv) print("exit successfully") diff --git a/scripts/maintenance/utilities.py b/scripts/maintenance/utilities.py new file mode 100644 index 00000000..886410c2 --- /dev/null +++ b/scripts/maintenance/utilities.py @@ -0,0 +1,89 @@ +import MySQLdb +import re +import configparser + +def get_cursor(): + host = 'tux.uthsc.edu' + user = 'webqtlout' + passwd = 'webqtlout' + db = 'db_webqtl' + con = MySQLdb.Connect(db=db, host=host, user=user, passwd=passwd) + cursor = con.cursor() + return cursor, con + +def clearspaces(s, default=None): + if s: + s = re.sub('\s+', ' ', s) + s = s.strip() + return s + else: + return default + +def to_dic(keys, values): + dic = {} + for i in range(len(keys)): + key = keys[i] + value = values[i] + dic[key] = value + return dic + +def overlap(dic1, dic2): + keys = [] + values1 = [] + values2 = [] + for key in dic1.keys(): + if key in dic2: + value1 = dic1[key] + value2 = dic2[key] + if value1 and value2: + keys.append(key) + values1.append(value1) + values2.append(value2) + return keys, values1, values2 + +def to_db_string(s, default): + if s: + s = s.strip() + if len(s) == 0: + return default + elif s == 'x': + return default + else: + return s + else: + return default + +def to_db_float(s, default): + if s: + s = s.strip() + if len(s) == 0: + return default + elif s == 'x': + return default + else: + try: + return float(s) + except: + return default + else: + return default + +def to_db_int(s, default): + if s: + s = s.strip() + if len(s) == 0: + return default + elif s == 'x': + return default + else: + try: + return int(s) + except: + return default + else: + return default + +def get_config(configfile): + config = configparser.ConfigParser() + config.read(configfile) + return config |