diff options
author | BonfaceKilz | 2021-11-25 20:52:14 +0300 |
---|---|---|
committer | BonfaceKilz | 2021-11-25 20:52:14 +0300 |
commit | a1516993c7f6dc608f75ba42cb27b983e0c5c330 (patch) | |
tree | fe0ac9871df6e0bad19a8aced77eee2dd7e303a2 /gn3/db | |
parent | c46f95ee85142f2a132db7e1a35691bb525061e0 (diff) | |
download | genenetwork3-a1516993c7f6dc608f75ba42cb27b983e0c5c330.tar.gz |
db: traits: Support additions and deletions from csv file
Diffstat (limited to 'gn3/db')
-rw-r--r-- | gn3/db/traits.py | 238 |
1 files changed, 193 insertions, 45 deletions
diff --git a/gn3/db/traits.py b/gn3/db/traits.py index ebb7e3c..75de4f4 100644 --- a/gn3/db/traits.py +++ b/gn3/db/traits.py @@ -1,5 +1,6 @@ """This class contains functions relating to trait data manipulation""" import os +import MySQLdb from functools import reduce from typing import Any, Dict, Union, Sequence @@ -76,16 +77,15 @@ def export_trait_data( return reduce(__exporter, samplelist, tuple()) + def get_trait_csv_sample_data(conn: Any, trait_name: int, phenotype_id: int): """Fetch a trait and return it as a csv string""" - def __float_strip(n): if str(n)[-2:] == ".0": return str(int(n)) return str(n) - sql = ("SELECT DISTINCT Strain.Id, PublishData.Id, Strain.Name, " - "PublishData.value, " + sql = ("SELECT DISTINCT Strain.Name, PublishData.value, " "PublishSE.error, NStrain.count FROM " "(PublishData, Strain, PublishXRef, PublishFreeze) " "LEFT JOIN PublishSE ON " @@ -97,65 +97,188 @@ def get_trait_csv_sample_data(conn: Any, "PublishData.Id = PublishXRef.DataId AND " "PublishXRef.Id = %s AND PublishXRef.PhenotypeId = %s " "AND PublishData.StrainId = Strain.Id Order BY Strain.Name") - csv_data = ["Strain Id,Strain Name,Value,SE,Count"] - publishdata_id = "" + csv_data = ["Strain Name,Value,SE,Count"] with conn.cursor() as cursor: cursor.execute(sql, (trait_name, phenotype_id,)) for record in cursor.fetchall(): - (strain_id, publishdata_id, - strain_name, value, error, count) = record + (strain_name, value, error, count) = record csv_data.append( ",".join([__float_strip(val) if val else "x" - for val in (strain_id, strain_name, - value, error, count)])) - return f"# Publish Data Id: {publishdata_id}\n" + "\n".join(csv_data) + for val in (strain_name, value, error, count)])) + return "\n".join(csv_data) def update_sample_data(conn: Any, + trait_name: str, strain_name: str, - strain_id: int, - publish_data_id: int, + phenotype_id: int, value: Union[int, float, str], error: Union[int, float, str], count: Union[int, str]): """Given the right parameters, update sample-data from the relevant table.""" - # pylint: disable=[R0913, R0914, C0103] - STRAIN_ID_SQL: str = "UPDATE Strain SET Name = %s WHERE Id = %s" - PUBLISH_DATA_SQL: str = ("UPDATE PublishData SET value = %s " - "WHERE StrainId = %s AND Id = %s") - PUBLISH_SE_SQL: str = ("UPDATE PublishSE SET error = %s " - "WHERE StrainId = %s AND DataId = %s") - N_STRAIN_SQL: str = ("UPDATE NStrain SET count = %s " - "WHERE StrainId = %s AND DataId = %s") - - updated_strains: int = 0 + strain_id, data_id = "", "" + + with conn.cursor() as cursor: + cursor.execute( + ("SELECT Strain.Id, PublishData.Id FROM " + "(PublishData, Strain, PublishXRef, PublishFreeze) " + "LEFT JOIN PublishSE ON " + "(PublishSE.DataId = PublishData.Id AND " + "PublishSE.StrainId = PublishData.StrainId) " + "LEFT JOIN NStrain ON " + "(NStrain.DataId = PublishData.Id AND " + "NStrain.StrainId = PublishData.StrainId) " + "WHERE PublishXRef.InbredSetId = " + "PublishFreeze.InbredSetId AND " + "PublishData.Id = PublishXRef.DataId AND " + "PublishXRef.Id = %s AND " + "PublishXRef.PhenotypeId = %s " + "AND PublishData.StrainId = Strain.Id " + "AND Strain.Name = \"%s\"") % (trait_name, + phenotype_id, + str(strain_name))) + strain_id, data_id = cursor.fetchone() updated_published_data: int = 0 updated_se_data: int = 0 updated_n_strains: int = 0 with conn.cursor() as cursor: - # Update the Strains table - cursor.execute(STRAIN_ID_SQL, (strain_name, strain_id)) - updated_strains = cursor.rowcount # Update the PublishData table - cursor.execute(PUBLISH_DATA_SQL, + cursor.execute(("UPDATE PublishData SET value = %s " + "WHERE StrainId = %s AND Id = %s"), (None if value == "x" else value, - strain_id, publish_data_id)) + strain_id, data_id)) updated_published_data = cursor.rowcount + # Update the PublishSE table - cursor.execute(PUBLISH_SE_SQL, + cursor.execute(("UPDATE PublishSE SET error = %s " + "WHERE StrainId = %s AND DataId = %s"), (None if error == "x" else error, - strain_id, publish_data_id)) + strain_id, data_id)) updated_se_data = cursor.rowcount + # Update the NStrain table - cursor.execute(N_STRAIN_SQL, + cursor.execute(("UPDATE NStrain SET count = %s " + "WHERE StrainId = %s AND DataId = %s"), (None if count == "x" else count, - strain_id, publish_data_id)) + strain_id, data_id)) updated_n_strains = cursor.rowcount - return (updated_strains, updated_published_data, + return (updated_published_data, updated_se_data, updated_n_strains) + +def delete_sample_data(conn: Any, + trait_name: str, + strain_name: str, + phenotype_id: int): + """Given the right parameters, delete sample-data from the relevant + table.""" + strain_id, data_id = "", "" + + deleted_published_data: int = 0 + deleted_se_data: int = 0 + deleted_n_strains: int = 0 + + with conn.cursor() as cursor: + # Delete the PublishData table + try: + cursor.execute( + ("SELECT Strain.Id, PublishData.Id FROM " + "(PublishData, Strain, PublishXRef, PublishFreeze) " + "LEFT JOIN PublishSE ON " + "(PublishSE.DataId = PublishData.Id AND " + "PublishSE.StrainId = PublishData.StrainId) " + "LEFT JOIN NStrain ON " + "(NStrain.DataId = PublishData.Id AND " + "NStrain.StrainId = PublishData.StrainId) " + "WHERE PublishXRef.InbredSetId = " + "PublishFreeze.InbredSetId AND " + "PublishData.Id = PublishXRef.DataId AND " + "PublishXRef.Id = %s AND " + "PublishXRef.PhenotypeId = %s " + "AND PublishData.StrainId = Strain.Id " + "AND Strain.Name = \"%s\"") % (trait_name, + phenotype_id, + str(strain_name))) + strain_id, data_id = cursor.fetchone() + + cursor.execute(("DELETE FROM PublishData " + "WHERE StrainId = %s AND Id = %s") + % (strain_id, data_id)) + deleted_published_data = cursor.rowcount + + # Delete the PublishSE table + cursor.execute(("DELETE FROM PublishSE " + "WHERE StrainId = %s AND DataId = %s") % + (strain_id, data_id)) + deleted_se_data = cursor.rowcount + + # Delete the NStrain table + cursor.execute(("DELETE FROM NStrain " + "WHERE StrainId = %s AND DataId = %s" % + (strain_id, data_id))) + deleted_n_strains = cursor.rowcount + except Exception as e: + conn.rollback() + raise MySQLdb.Error + conn.commit() + cursor.close() + cursor.close() + + return (deleted_published_data, + deleted_se_data, deleted_n_strains) + + +def insert_sample_data(conn: Any, + trait_name: str, + strain_name: str, + phenotype_id: int, + value: Union[int, float, str], + error: Union[int, float, str], + count: Union[int, str]): + """Given the right parameters, insert sample-data to the relevant table. + + """ + + inserted_published_data, inserted_se_data, inserted_n_strains = 0, 0, 0 + with conn.cursor() as cursor: + try: + cursor.execute("SELECT DataId FROM PublishXRef WHERE Id = %s AND " + "PhenotypeId = %s", (trait_name, phenotype_id)) + data_id = cursor.fetchone() + + cursor.execute("SELECT Id FROM Strain WHERE Name = %s", + (strain_name,)) + strain_id = cursor.fetchone() + + # Insert the PublishData table + cursor.execute(("INSERT INTO PublishData (Id, StrainId, value)" + "VALUES (%s, %s, %s)"), + (data_id, strain_id, value)) + inserted_published_data = cursor.rowcount + + # Insert into the PublishSE table if error is specified + if error and error != "x": + cursor.execute(("INSERT INTO PublishSE (StrainId, DataId, " + " error) VALUES (%s, %s, %s)") % + (strain_id, data_id, error)) + inserted_se_data = cursor.rowcount + + # Insert into the NStrain table + if count and count != "x": + cursor.execute(("INSERT INTO NStrain " + "(StrainId, DataId, error) " + "VALUES (%s, %s, %s)") % + (strain_id, data_id, count)) + inserted_n_strains = cursor.rowcount + except Exception as e: + conn.rollback() + raise MySQLdb.Error + return (inserted_published_data, + inserted_se_data, inserted_n_strains) + + def retrieve_publish_trait_info(trait_data_source: Dict[str, Any], conn: Any): """Retrieve trait information for type `Publish` traits. @@ -195,11 +318,12 @@ def retrieve_publish_trait_info(trait_data_source: Dict[str, Any], conn: Any): cursor.execute( query, { - k:v for k, v in trait_data_source.items() + k: v for k, v in trait_data_source.items() if k in ["trait_name", "trait_dataset_id"] }) return dict(zip([k.lower() for k in keys], cursor.fetchone())) + def set_confidential_field(trait_type, trait_info): """Post processing function for 'Publish' trait types. @@ -212,6 +336,7 @@ def set_confidential_field(trait_type, trait_info): and not trait_info.get("pubmed_id", None)) else 0} return trait_info + def retrieve_probeset_trait_info(trait_data_source: Dict[str, Any], conn: Any): """Retrieve trait information for type `ProbeSet` traits. @@ -239,11 +364,12 @@ def retrieve_probeset_trait_info(trait_data_source: Dict[str, Any], conn: Any): cursor.execute( query, { - k:v for k, v in trait_data_source.items() + k: v for k, v in trait_data_source.items() if k in ["trait_name", "trait_dataset_name"] }) return dict(zip(keys, cursor.fetchone())) + def retrieve_geno_trait_info(trait_data_source: Dict[str, Any], conn: Any): """Retrieve trait information for type `Geno` traits. @@ -263,11 +389,12 @@ def retrieve_geno_trait_info(trait_data_source: Dict[str, Any], conn: Any): cursor.execute( query, { - k:v for k, v in trait_data_source.items() + k: v for k, v in trait_data_source.items() if k in ["trait_name", "trait_dataset_name"] }) return dict(zip(keys, cursor.fetchone())) + def retrieve_temp_trait_info(trait_data_source: Dict[str, Any], conn: Any): """Retrieve trait information for type `Temp` traits. @@ -280,11 +407,12 @@ def retrieve_temp_trait_info(trait_data_source: Dict[str, Any], conn: Any): cursor.execute( query, { - k:v for k, v in trait_data_source.items() + k: v for k, v in trait_data_source.items() if k in ["trait_name"] }) return dict(zip(keys, cursor.fetchone())) + def set_haveinfo_field(trait_info): """ Common postprocessing function for all trait types. @@ -292,6 +420,7 @@ def set_haveinfo_field(trait_info): Sets the value for the 'haveinfo' field.""" return {**trait_info, "haveinfo": 1 if trait_info else 0} + def set_homologene_id_field_probeset(trait_info, conn): """ Postprocessing function for 'ProbeSet' traits. @@ -307,7 +436,7 @@ def set_homologene_id_field_probeset(trait_info, conn): cursor.execute( query, { - k:v for k, v in trait_info.items() + k: v for k, v in trait_info.items() if k in ["geneid", "group"] }) res = cursor.fetchone() @@ -315,12 +444,13 @@ def set_homologene_id_field_probeset(trait_info, conn): return {**trait_info, "homologeneid": res[0]} return {**trait_info, "homologeneid": None} + def set_homologene_id_field(trait_type, trait_info, conn): """ Common postprocessing function for all trait types. Sets the value for the 'homologene' key.""" - set_to_null = lambda ti: {**ti, "homologeneid": None} + def set_to_null(ti): return {**ti, "homologeneid": None} functions_table = { "Temp": set_to_null, "Geno": set_to_null, @@ -329,6 +459,7 @@ def set_homologene_id_field(trait_type, trait_info, conn): } return functions_table[trait_type](trait_info) + def load_publish_qtl_info(trait_info, conn): """ Load extra QTL information for `Publish` traits @@ -349,6 +480,7 @@ def load_publish_qtl_info(trait_info, conn): return dict(zip(["locus", "lrs", "additive"], cursor.fetchone())) return {"locus": "", "lrs": "", "additive": ""} + def load_probeset_qtl_info(trait_info, conn): """ Load extra QTL information for `ProbeSet` traits @@ -371,6 +503,7 @@ def load_probeset_qtl_info(trait_info, conn): ["locus", "lrs", "pvalue", "mean", "additive"], cursor.fetchone())) return {"locus": "", "lrs": "", "pvalue": "", "mean": "", "additive": ""} + def load_qtl_info(qtl, trait_type, trait_info, conn): """ Load extra QTL information for traits @@ -399,6 +532,7 @@ def load_qtl_info(qtl, trait_type, trait_info, conn): return qtl_info_functions[trait_type](trait_info, conn) + def build_trait_name(trait_fullname): """ Initialises the trait's name, and other values from the search data provided @@ -425,6 +559,7 @@ def build_trait_name(trait_fullname): "cellid": name_parts[2] if len(name_parts) == 3 else "" } + def retrieve_probeset_sequence(trait, conn): """ Retrieve a 'ProbeSet' trait's sequence information @@ -446,6 +581,7 @@ def retrieve_probeset_sequence(trait, conn): seq = cursor.fetchone() return {**trait, "sequence": seq[0] if seq else ""} + def retrieve_trait_info( threshold: int, trait_full_name: str, conn: Any, qtl=None): @@ -501,6 +637,7 @@ def retrieve_trait_info( } return trait_info + def retrieve_temp_trait_data(trait_info: dict, conn: Any): """ Retrieve trait data for `Temp` traits. @@ -520,9 +657,10 @@ def retrieve_temp_trait_data(trait_info: dict, conn: Any): {"trait_name": trait_info["trait_name"]}) return [dict(zip( ["sample_name", "value", "se_error", "nstrain", "id"], row)) - for row in cursor.fetchall()] + for row in cursor.fetchall()] return [] + def retrieve_species_id(group, conn: Any): """ Retrieve a species id given the Group value @@ -534,6 +672,7 @@ def retrieve_species_id(group, conn: Any): return cursor.fetchone()[0] return None + def retrieve_geno_trait_data(trait_info: Dict, conn: Any): """ Retrieve trait data for `Geno` traits. @@ -559,9 +698,10 @@ def retrieve_geno_trait_data(trait_info: Dict, conn: Any): trait_info["db"]["group"], conn)}) return [dict(zip( ["sample_name", "value", "se_error", "id"], row)) - for row in cursor.fetchall()] + for row in cursor.fetchall()] return [] + def retrieve_publish_trait_data(trait_info: Dict, conn: Any): """ Retrieve trait data for `Publish` traits. @@ -590,9 +730,10 @@ def retrieve_publish_trait_data(trait_info: Dict, conn: Any): "dataset_id": trait_info["db"]["dataset_id"]}) return [dict(zip( ["sample_name", "value", "se_error", "nstrain", "id"], row)) - for row in cursor.fetchall()] + for row in cursor.fetchall()] return [] + def retrieve_cellid_trait_data(trait_info: Dict, conn: Any): """ Retrieve trait data for `Probe Data` types. @@ -623,9 +764,10 @@ def retrieve_cellid_trait_data(trait_info: Dict, conn: Any): "dataset_id": trait_info["db"]["dataset_id"]}) return [dict(zip( ["sample_name", "value", "se_error", "id"], row)) - for row in cursor.fetchall()] + for row in cursor.fetchall()] return [] + def retrieve_probeset_trait_data(trait_info: Dict, conn: Any): """ Retrieve trait data for `ProbeSet` traits. @@ -652,9 +794,10 @@ def retrieve_probeset_trait_data(trait_info: Dict, conn: Any): "dataset_name": trait_info["db"]["dataset_name"]}) return [dict(zip( ["sample_name", "value", "se_error", "id"], row)) - for row in cursor.fetchall()] + for row in cursor.fetchall()] return [] + def with_samplelist_data_setup(samplelist: Sequence[str]): """ Build function that computes the trait data from provided list of samples. @@ -681,6 +824,7 @@ def with_samplelist_data_setup(samplelist: Sequence[str]): return None return setup_fn + def without_samplelist_data_setup(): """ Build function that computes the trait data. @@ -701,6 +845,7 @@ def without_samplelist_data_setup(): return None return setup_fn + def retrieve_trait_data(trait: dict, conn: Any, samplelist: Sequence[str] = tuple()): """ Retrieve trait data @@ -740,15 +885,17 @@ def retrieve_trait_data(trait: dict, conn: Any, samplelist: Sequence[str] = tupl "data": dict(map( lambda x: ( x["sample_name"], - {k:v for k, v in x.items() if x != "sample_name"}), + {k: v for k, v in x.items() if x != "sample_name"}), data))} return {} + def generate_traits_filename(base_path: str = TMPDIR): """Generate a unique filename for use with generated traits files.""" return "{}/traits_test_file_{}.txt".format( os.path.abspath(base_path), random_string(10)) + def export_informative(trait_data: dict, inc_var: bool = False) -> tuple: """ Export informative strain @@ -770,5 +917,6 @@ def export_informative(trait_data: dict, inc_var: bool = False) -> tuple: return acc return reduce( __exporter__, - filter(lambda td: td["value"] is not None, trait_data["data"].values()), + filter(lambda td: td["value"] is not None, + trait_data["data"].values()), (tuple(), tuple(), tuple())) |