diff options
Diffstat (limited to 'wqflask')
-rw-r--r-- | wqflask/base/trait.py | 94 |
1 files changed, 80 insertions, 14 deletions
diff --git a/wqflask/base/trait.py b/wqflask/base/trait.py index b133bf21..405c4ebf 100644 --- a/wqflask/base/trait.py +++ b/wqflask/base/trait.py @@ -1,9 +1,11 @@ from __future__ import absolute_import, division, print_function +import os import string import resource import codecs import requests +import random from base import webqtlConfig from base.webqtlCaseData import webqtlCaseData @@ -12,8 +14,8 @@ from db import webqtlDatabaseFunction from utility import webqtlUtil from utility import hmac from utility.authentication_tools import check_resource_availability -from utility.tools import GN2_BASE_URL -from utility.redis_tools import get_redis_conn +from utility.tools import GN2_BASE_URL, GN_VERSION +from utility.redis_tools import get_redis_conn, get_resource_id, get_resource_info Redis = get_redis_conn() from wqflask import app @@ -22,7 +24,7 @@ import simplejson as json from MySQLdb import escape_string as escape from pprint import pformat as pf -from flask import Flask, g, request, url_for, redirect +from flask import Flask, g, request, url_for, redirect, make_response, render_template from utility.logger import getLogger logger = getLogger(__name__ ) @@ -45,7 +47,10 @@ def create_trait(**kw): permitted = check_resource_availability(dataset) if permitted: - return GeneralTrait(**kw) + the_trait = GeneralTrait(**kw) + if the_trait.dataset.type != "Temp": + the_trait = retrieve_trait_info(the_trait, the_trait.dataset, get_qtl_info=kw.get('get_qtl_info')) + return the_trait else: return None @@ -99,9 +104,6 @@ class GeneralTrait(object): # Todo: These two lines are necessary most of the time, but perhaps not all of the time # So we could add a simple if statement to short-circuit this if necessary - if self.dataset.type != "Temp": - self = retrieve_trait_info(self, self.dataset, get_qtl_info=get_qtl_info) - if get_sample_info != False: self = retrieve_sample_data(self, self.dataset) @@ -373,17 +375,15 @@ def jsonable_table_row(trait, dataset_name, index): else: return dict() + def retrieve_trait_info(trait, dataset, get_qtl_info=False): assert dataset, "Dataset doesn't exist" + resource_id = get_resource_id(dataset, trait.name) if dataset.type == 'Publish': - resource_id = hmac.hmac_creation("{}:{}:{}".format('dataset-publish', dataset.id, trait.name)) the_url = "http://localhost:8080/run-action?resource={}&user={}&branch=data&action=view".format(resource_id, g.user_session.user_id) - elif dataset.type == 'ProbeSet': - resource_id = hmac.hmac_creation("{}:{}".format('dataset-probeset', dataset.id)) - the_url = "http://localhost:8080/run-action?resource={}&user={}&branch=data&action=view&trait={}".format(resource_id, g.user_session.user_id, trait.name) else: - resource_id = hmac.hmac_creation("{}:{}".format('dataset-geno', dataset.id)) + the_url = "http://localhost:8080/run-action?resource={}&user={}&branch=data&action=view&trait={}".format(resource_id, g.user_session.user_id, trait.name) try: @@ -394,11 +394,77 @@ def retrieve_trait_info(trait, dataset, get_qtl_info=False): except: resource_info = get_resource_info(resource_id) default_permissions = resource_info['default_mask']['data'] - if 'view' not in default_persmissions: + if 'view' not in default_permissions: trait.view = False return trait - trait_info = json.loads(response) + if dataset.type == 'Publish': + query = """ + SELECT + PublishXRef.Id, Publication.PubMed_ID, + Phenotype.Pre_publication_description, Phenotype.Post_publication_description, Phenotype.Original_description, + Phenotype.Pre_publication_abbreviation, Phenotype.Post_publication_abbreviation, + Phenotype.Lab_code, Phenotype.Submitter, Phenotype.Owner, Phenotype.Authorized_Users, + Publication.Authors, Publication.Title, Publication.Abstract, + Publication.Journal, Publication.Volume, Publication.Pages, + Publication.Month, Publication.Year, PublishXRef.Sequence, + Phenotype.Units, PublishXRef.comments + FROM + PublishXRef, Publication, Phenotype, PublishFreeze + WHERE + PublishXRef.Id = %s AND + Phenotype.Id = PublishXRef.PhenotypeId AND + Publication.Id = PublishXRef.PublicationId AND + PublishXRef.InbredSetId = PublishFreeze.InbredSetId AND + PublishFreeze.Id = %s + """ % (trait.name, dataset.id) + + logger.sql(query) + trait_info = g.db.execute(query).fetchone() + + + #XZ, 05/08/2009: Xiaodong add this block to use ProbeSet.Id to find the probeset instead of just using ProbeSet.Name + #XZ, 05/08/2009: to avoid the problem of same probeset name from different platforms. + elif dataset.type == 'ProbeSet': + display_fields_string = ', ProbeSet.'.join(dataset.display_fields) + display_fields_string = 'ProbeSet.' + display_fields_string + query = """ + SELECT %s + FROM ProbeSet, ProbeSetFreeze, ProbeSetXRef + WHERE + ProbeSetXRef.ProbeSetFreezeId = ProbeSetFreeze.Id AND + ProbeSetXRef.ProbeSetId = ProbeSet.Id AND + ProbeSetFreeze.Name = '%s' AND + ProbeSet.Name = '%s' + """ % (escape(display_fields_string), + escape(dataset.name), + escape(str(trait.name))) + logger.sql(query) + trait_info = g.db.execute(query).fetchone() + #XZ, 05/08/2009: We also should use Geno.Id to find marker instead of just using Geno.Name + # to avoid the problem of same marker name from different species. + elif dataset.type == 'Geno': + display_fields_string = string.join(dataset.display_fields,',Geno.') + display_fields_string = 'Geno.' + display_fields_string + query = """ + SELECT %s + FROM Geno, GenoFreeze, GenoXRef + WHERE + GenoXRef.GenoFreezeId = GenoFreeze.Id AND + GenoXRef.GenoId = Geno.Id AND + GenoFreeze.Name = '%s' AND + Geno.Name = '%s' + """ % (escape(display_fields_string), + escape(dataset.name), + escape(trait.name)) + logger.sql(query) + trait_info = g.db.execute(query).fetchone() + else: #Temp type + query = """SELECT %s FROM %s WHERE Name = %s""" + logger.sql(query) + trait_info = g.db.execute(query, + (string.join(dataset.display_fields,','), + dataset.type, trait.name)).fetchone() if trait_info: trait.haveinfo = True |