Reorganise the database code

Issue:
https://github.com/genenetwork/gn-gemtext-threads/blob/main/topics/gn1-migration-to-gn2/clustering.gmi

* Reorganise the code to separate the datasets from the traits, and to more
  closely conform to the same flow as that in GN1

2 files changed, 324 insertions, 98 deletions

diff --git a/gn3/db/datasets.py b/gn3/db/datasets.py
new file mode 100644
index 0000000..3ad50f6
--- /dev/null
+++ b/gn3/db/datasets.py
@@ -0,0 +1,251 @@
+from typing import Any, Dict, Union
+
+def retrieve_probeset_trait_dataset_name(
+        threshold: int, name: str, connection: Any):
+    query = (
+        "SELECT Id, Name, FullName, ShortName, DataScale "
+        "FROM ProbeSetFreeze "
+        "WHERE "
+        "public > %(threshold)s "
+        "AND "
+        "(Name = %(name)s OR FullName = %(name)s OR ShortName = %(name)s)")
+    with connection.cursor() as cursor:
+        cursor.execute(
+            query,
+            {
+                "threshold": threshold,
+                "name": name
+            })
+        return dict(zip(
+            ["dataset_id", "dataset_name", "dataset_fullname",
+             "dataset_shortname", "dataset_datascale"],
+            cursor.fetchone))
+
+def retrieve_publish_trait_dataset_name(threshold: int, name: str, connection: Any):
+    query = (
+        "SELECT Id, Name, FullName, ShortName "
+        "FROM PublishFreeze "
+        "WHERE "
+        "public > %(threshold)s "
+        "AND "
+        "(Name = %(name)s OR FullName = %(name)s OR ShortName = %(name)s)")
+    with connection.cursor() as cursor:
+        cursor.execute(
+            query,
+            {
+                "threshold": threshold,
+                "name": name
+            })
+        return dict(zip(
+            ["dataset_id", "dataset_name", "dataset_fullname",
+             "dataset_shortname"],
+            cursor.fetchone))
+
+def retrieve_geno_trait_dataset_name(threshold: int, name: str, connection: Any):
+    query = (
+        "SELECT Id, Name, FullName, ShortName "
+        "FROM GenoFreeze "
+        "WHERE "
+        "public > %(threshold)s "
+        "AND "
+        "(Name = %(name)s OR FullName = %(name)s OR ShortName = %(name)s)")
+    with connection.cursor() as cursor:
+        cursor.execute(
+            query,
+            {
+                "threshold": threshold,
+                "name": name
+            })
+        return dict(zip(
+            ["dataset_id", "dataset_name", "dataset_fullname",
+             "dataset_shortname"],
+            cursor.fetchone))
+
+def retrieve_temp_trait_dataset_name(threshold: int, name: str, connection: Any):
+    query = (
+        "SELECT Id, Name, FullName, ShortName "
+        "FROM TempFreeze "
+        "WHERE "
+        "public > %(threshold)s "
+        "AND "
+        "(Name = %(name)s OR FullName = %(name)s OR ShortName = %(name)s)")
+    with connection.cursor() as cursor:
+        cursor.execute(
+            query,
+            {
+                "threshold": threshold,
+                "name": name
+            })
+        return dict(zip(
+            ["dataset_id", "dataset_name", "dataset_fullname",
+             "dataset_shortname"],
+            cursor.fetchone))
+
+def retrieve_dataset_name(
+        trait_type: str, threshold: int, trait_name: str, dataset_name: str,
+        conn: Any):
+    """
+    Retrieve the name of a trait given the trait's name
+
+    This is extracted from the `webqtlDataset.retrieveName` function as is
+    implemented at
+    https://github.com/genenetwork/genenetwork1/blob/master/web/webqtl/base/webqtlDataset.py#L140-L169
+    """
+    fn_map = {
+        "ProbeSet": retrieve_probeset_trait_dataset_name,
+        "Publish": retrieve_publish_trait_dataset_name,
+        "Geno": retrieve_geno_trait_dataset_name,
+        "Temp": retrieve_temp_trait_dataset_name}
+    if trait_type == "Temp":
+        return retrieve_temp_trait_dataset_name(threshold, trait_name, conn)
+    return fn_map[trait_type](threshold, dataset_name, conn)
+
+
+def retrieve_geno_riset_fields(name, conn):
+    """
+    Retrieve the RISet, and RISetID values for various Geno trait types.
+    """
+    query = (
+        "SELECT InbredSet.Name, InbredSet.Id "
+        "FROM InbredSet, GenoFreeze "
+        "WHERE GenoFreeze.InbredSetId = InbredSet.Id "
+        "AND GenoFreeze.Name = %(name)s")
+    with conn.cursor() as cursor:
+        cursor.execute(query, {"name": name})
+        return dict(zip(["riset", "risetid"], cursor.fetchone()))
+    return {}
+
+def retrieve_publish_riset_fields(name, conn):
+    """
+    Retrieve the RISet, and RISetID values for various Publish trait types.
+    """
+    query = (
+        "SELECT InbredSet.Name, InbredSet.Id "
+        "FROM InbredSet, PublishFreeze "
+        "WHERE PublishFreeze.InbredSetId = InbredSet.Id "
+        "AND PublishFreeze.Name = %(name)s")
+    with conn.cursor() as cursor:
+        cursor.execute(query, {"name": name})
+        return dict(zip(["riset", "risetid"], cursor.fetchone()))
+    return {}
+
+def retrieve_probeset_riset_fields(name, conn):
+    """
+    Retrieve the RISet, and RISetID values for various ProbeSet trait types.
+    """
+    query = (
+        "SELECT InbredSet.Name, InbredSet.Id "
+        "FROM InbredSet, ProbeSetFreeze, ProbeFreeze "
+        "WHERE ProbeFreeze.InbredSetId = InbredSet.Id "
+        "AND ProbeFreeze.Id = ProbeSetFreeze.ProbeFreezeId "
+        "AND ProbeSetFreeze.Name = %(name)s")
+    with conn.cursor() as cursor:
+        cursor.execute(query, {"name": name})
+        return dict(zip(["riset", "risetid"], cursor.fetchone()))
+    return {}
+
+def retrieve_temp_riset_fields(name, conn):
+    query = (
+        "SELECT InbredSet.Name, InbredSet.Id "
+        "FROM InbredSet, Temp "
+        "WHERE Temp.InbredSetId = InbredSet.Id "
+        "AND Temp.Name = %(name)s")
+    with conn.cursor() as cursor:
+        cursor.execute(query, {"name": name})
+        return dict(zip(["riset", "risetid"], cursor.fetchone()))
+    return {}
+
+def retrieve_riset_fields(trait_type, trait_name, dataset_info, conn):
+    """
+    Retrieve the RISet, and RISetID values for various trait types.
+    """
+    riset_fns_map = {
+        "Geno": retrieve_geno_riset_fields,
+        "Publish": retrieve_publish_riset_fields,
+        "ProbeSet": retrieve_probeset_riset_fields
+    }
+
+    if trait_type == "Temp":
+        riset_info = retrieve_temp_riset_fields(trait_name, conn)
+    else:
+        riset_info = riset_fns_map[trait_type](dataset_info["dataset_name"], conn)
+
+    return {
+        **dataset_info,
+        **riset_info,
+        "riset": (
+            "BXD" if riset_info.get("riset") == "BXD300"
+            else riset_info.get("riset", ""))
+    }
+
+def retrieve_temp_trait_dataset():
+    return {
+        "searchfield": ["name", "description"],
+        "disfield": ["name", "description"],
+        "type": "Temp",
+        "dataset_id": 1,
+        "fullname": "Temporary Storage",
+        "shortname": "Temp"
+    }
+
+def retrieve_geno_trait_dataset():
+    return {
+        "searchfield": ["name","chr"],
+	"disfield": ["name","chr","mb", "source2", "sequence"],
+	"type": "Geno"
+    }
+
+def retrieve_publish_trait_dataset():
+    return {
+        "searchfield": [
+            "name", "post_publication_description", "abstract", "title",
+            "authors"],
+        "disfield": [
+            "name","pubmed_id", "pre_publication_description",
+            "post_publication_description", "original_description", 
+	    "pre_publication_abbreviation", "post_publication_abbreviation",
+	    "lab_code", "submitter", "owner", "authorized_users",
+	    "authors","title","abstract", "journal","volume","pages","month",
+	    "year","sequence", "units", "comments"],
+        "type": "Publish"
+    }
+
+def retrieve_probeset_trait_dataset():
+    return {
+        "searchfield": [
+            "name", "description", "probe_target_description", "symbol",
+            "alias", "genbankid", "unigeneid", "omim", "refseq_transcriptid",
+            "probe_set_specificity", "probe_set_blat_score"],
+	"disfield": [
+            "name", "symbol", "description", "probe_target_description", "chr",
+            "mb", "alias", "geneid", "genbankid", "unigeneid", "omim",
+            "refseq_transcriptid", "blatseq", "targetseq", "chipid", "comments",
+            "strand_probe", "strand_gene", "probe_set_target_region",
+            "proteinid", "probe_set_specificity", "probe_set_blat_score",
+            "probe_set_blat_mb_start", "probe_set_blat_mb_end",
+            "probe_set_strand", "probe_set_note_by_rw", "flag"],
+	"type": "ProbeSet"
+    }
+
+def retrieve_trait_dataset(trait_type, trait, threshold, conn):
+    dataset_fns = {
+        "Temp": retrieve_temp_trait_dataset,
+        "Geno": retrieve_geno_trait_dataset,
+        "Publish": retrieve_publish_trait_dataset,
+        "ProbeSet": retrieve_probeset_trait_dataset
+    }
+    dataset_name_info = {
+        "dataset_id": None,
+        "dataset_name": trait["db"]["dataset_name"],
+        **retrieve_dataset_name(
+        trait_type, threshold, trait["trait_name"], trait["db"]["dataset_name"],
+        conn)
+    }
+    riset = retrieve_riset_fields(
+        trait_type, trait["trait_name"], dataset_name_info, conn)
+    return {
+        "display_name": dataset_name_info["dataset_name"],
+        **dataset_name_info,
+        **dataset_fns[trait_type](),
+        **riset
+    }
diff --git a/gn3/db/traits.py b/gn3/db/traits.py
index 9f89510..85cccfa 100644
--- a/gn3/db/traits.py
+++ b/gn3/db/traits.py
@@ -1,6 +1,7 @@
 """This class contains functions relating to trait data manipulation"""
 from typing import Any, Dict, Union
 from gn3.function_helpers import compose
+from gn3.db.datasets import retrieve_trait_dataset
 
 
 def get_trait_csv_sample_data(conn: Any,
@@ -77,41 +78,6 @@ def update_sample_data(conn: Any,
     return (updated_strains, updated_published_data,
             updated_se_data, updated_n_strains)
 
-def retrieve_trait_dataset_name(
-        trait_type: str, threshold: int, name: str, connection: Any):
-    """
-    Retrieve the name of a trait given the trait's name
-
-    This is extracted from the `webqtlDataset.retrieveName` function as is
-    implemented at
-    https://github.com/genenetwork/genenetwork1/blob/master/web/webqtl/base/webqtlDataset.py#L140-L169
-    """
-    table_map = {
-        "ProbeSet": "ProbeSetFreeze",
-        "Publish": "PublishFreeze",
-        "Geno": "GenoFreeze",
-        "Temp": "TempFreeze"}
-    columns = "Id, Name, FullName, ShortName{}".format(
-        ", DataScale" if trait_type == "ProbeSet" else "")
-    query = (
-        "SELECT %(columns)s "
-        "FROM %(table)s "
-        "WHERE "
-        "public > %(threshold)s "
-        "AND "
-        "(Name = %(name)s OR FullName = %(name)s OR ShortName = %(name)s)")
-    with connection.cursor() as cursor:
-        cursor.execute(
-            query,
-            {
-                "table": table_map[trait_type],
-                "columns": columns,
-                "threshold": threshold,
-                "name": name
-            })
-        return cursor.fetchone()
-
-
 def retrieve_publish_trait_info(trait_data_source: Dict[str, Any], conn: Any):
     """Retrieve trait information for type `Publish` traits.
 
@@ -156,11 +122,11 @@ def retrieve_publish_trait_info(trait_data_source: Dict[str, Any], conn: Any):
             })
         return dict(zip([k.lower() for k in keys], cursor.fetchone()))
 
-def set_confidential_field(trait_info):
+def set_confidential_field(trait_type, trait_info):
     """Post processing function for 'Publish' trait types.
 
     It sets the value for the 'confidential' key."""
-    if trait_info["type"] == "Publish":
+    if trait_type == "Publish":
         return {
             **trait_info,
             "confidential": 1 if (
@@ -271,7 +237,7 @@ 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_info, conn):
+def set_homologene_id_field(trait_type, trait_info, conn):
     """
     Common postprocessing function for all trait types.
 
@@ -283,84 +249,83 @@ def set_homologene_id_field(trait_info, conn):
         "Publish": set_to_null,
         "ProbeSet": lambda ti: set_homologene_id_field_probeset(ti, conn)
     }
-    return functions_table[trait_info["type"]](trait_info)
+    return functions_table[trait_type](trait_info)
 
-def set_geno_riset_fields(name, conn):
-    """
-    Retrieve the RISet, and RISetID values for various Geno trait types.
-    """
+def load_publish_qtl_info(trait_info, conn):
     query = (
-        "SELECT InbredSet.Name, InbredSet.Id "
-        "FROM InbredSet, GenoFreeze "
-        "WHERE GenoFreeze.InbredSetId = InbredSet.Id "
-        "AND GenoFreeze.Name = %(name)s")
+        "SELECT PublishXRef.Locus, PublishXRef.LRS, PublishXRef.additive "
+        "FROM PublishXRef, PublishFreeze "
+        "WHERE PublishXRef.Id = %(trait_name)s "
+        "AND PublishXRef.InbredSetId = PublishFreeze.InbredSetId "
+        "AND PublishFreeze.Id = %(dataset_id)s")
     with conn.cursor() as cursor:
-        return cursor.execute(query, {"name": name})
+        cursor.execute()
+        return dict(zip(["locus", "lrs", "additive"], cursor.fetchone()))
+    return {"locus": "", "lrs": "", "additive": ""}
 
-def set_publish_riset_fields(name, conn):
-    """
-    Retrieve the RISet, and RISetID values for various Publish trait types.
-    """
+def load_probeset_qtl_info(trait_info, conn):
     query = (
-        "SELECT InbredSet.Name, InbredSet.Id "
-        "FROM InbredSet, PublishFreeze "
-        "WHERE PublishFreeze.InbredSetId = InbredSet.Id "
-        "AND PublishFreeze.Name = %(name)s")
+        "SELECT ProbeSetXRef.Locus, ProbeSetXRef.LRS, ProbeSetXRef.pValue, "
+        "ProbeSetXRef.mean, ProbeSetXRef.additive "
+        "FROM ProbeSetXRef, ProbeSet "
+        "WHERE ProbeSetXRef.ProbeSetId = ProbeSet.Id "
+        " AND ProbeSet.Name = %(trait_name)s "
+        "AND ProbeSetXRef.ProbeSetFreezeId = %(dataset_id)s")
     with conn.cursor() as cursor:
-        return cursor.execute(query, {"name": name})
+        cursor.execute()
+        return dict(zip(
+            ["locus", "lrs", "pvalue", "mean", "additive"], cursor.fetchone()))
+    return {"locus": "", "lrs": "", "pvalue": "", "mean": "", "additive": ""}
 
-def set_probeset_riset_fields(name, conn):
-    """
-    Retrieve the RISet, and RISetID values for various ProbeSet trait types.
-    """
-    query = (
-        "SELECT InbredSet.Name, InbredSet.Id "
-        "FROM InbredSet, ProbeSetFreeze, ProbeFreeze "
-        "WHERE ProbeFreeze.InbredSetId = InbredSet.Id "
-        "AND ProbeFreeze.Id = ProbeSetFreeze.ProbeFreezeId "
-        "AND ProbeSetFreeze.Name = %(name)s")
-    with conn.cursor() as cursor:
-        return cursor.execute(query, {"name": name})
-
-def set_riset_fields(trait_info, conn):
-    """
-    Retrieve the RISet, and RISetID values for various trait types.
-    """
-    riset_functions_map = {
-        "Temp": lambda ti, con: (None, None),
-        "Geno": set_geno_riset_fields,
-        "Publish": set_publish_riset_fields,
-        "ProbeSet": set_probeset_riset_fields
+def load_qtl_info(qtl, trait_type, trait_info, conn):
+    if not qtl:
+        return trait_info
+    qtl_info_functions = {
+        "Publish": load_publish_qtl_info,
+        "ProbeSet": load_probeset_qtl_info
     }
-    if not trait_info.get("haveinfo", None):
+    if trait_inf["name"] not in qtl_info_functions.keys():
         return trait_info
 
-    riset, riid = riset_functions_map[trait_info["type"]](
-        trait_info["name"], conn)
-    return {
-        **trait_info, "risetid": riid,
-        "riset": "BXD" if riset == "BXD300" else riset}
+    return qtl_info_functions[trait_type](trait_info, conn)
 
 def build_trait_name(trait_fullname):
     name_parts = trait_fullname.split("::")
     assert len(name_parts) >= 2, "Name format error"
     return {
-        "trait_db": name_parts[0],
+        "db": {"dataset_name": name_parts[0]},
         "trait_fullname": trait_fullname,
         "trait_name": name_parts[1],
         "cellid": name_parts[2] if len(name_parts) == 3 else ""
     }
 
+def retrieve_probeset_sequence(trait, conn):
+    query = (
+        "SELECT ProbeSet.BlatSeq "
+        "FROM ProbeSet, ProbeSetFreeze, ProbeSetXRef "
+        "WHERE ProbeSet.Id=ProbeSetXRef.ProbeSetId "
+        "AND ProbeSetFreeze.Id = ProbeSetXRef.ProbeSetFreezeId "
+        "AND ProbeSet.Name = %(trait_name)s "
+        "AND ProbeSetFreeze.Name = %(dataset_name)s")
+    with conn.cursor() as cursor:
+        cursor.execute(
+            query,
+            {
+                "trait_name": trait["trait_name"],
+                "dataset_name": trait["db"]["dataset_name"]
+            })
+        seq = cursor.fetchone()
+        return {**trait, "sequence": seq[0] if seq else ""}
+
 def retrieve_trait_info(
-        trait_type: str, trait_full_name: str, trait_dataset_id: int,
-        trait_dataset_name: str, conn: Any, qtl=None):
+        trait_type: str, threshold: int, trait_full_name: str, conn: Any,
+        qtl=None):
     """Retrieves the trait information.
 
     https://github.com/genenetwork/genenetwork1/blob/master/web/webqtl/base/webqtlTrait.py#L397-L456
 
     This function, or the dependent functions, might be incomplete as they are
     currently."""
-    # pylint: disable=[R0913]
     trait = build_trait_name(trait_full_name)
     trait_info_function_table = {
         "Publish": retrieve_publish_trait_info,
@@ -370,15 +335,19 @@ def retrieve_trait_info(
     }
 
     common_post_processing_fn = compose(
-        lambda ti: set_riset_fields(ti, conn),
-        lambda ti: set_homologene_id_field(ti, conn),
-        lambda ti: {"type": trait_type, **ti},
-        lambda ti: {**ti, **trait},
+        lambda ti: load_qtl_info(qtl, trait_type, ti, conn),
+        lambda ti: set_homologene_id_field(trait_type, ti, conn),
+        lambda ti: {"trait_type": trait_type, **ti},
+        lambda ti: {**trait, **ti},
         set_haveinfo_field)
 
     trait_post_processing_functions_table = {
-        "Publish": compose(set_confidential_field, common_post_processing_fn),
-        "ProbeSet": compose(common_post_processing_fn),
+        "Publish": compose(
+            lambda ti: set_confidential_field(trait_type, ti),
+            common_post_processing_fn),
+        "ProbeSet": compose(
+            lambda ti: retrieve_probeset_sequence(ti, conn),
+            common_post_processing_fn),
         "Geno": common_post_processing_fn,
         "Temp": common_post_processing_fn
     }
@@ -387,10 +356,16 @@ def retrieve_trait_info(
         trait_post_processing_functions_table[trait_type],
         trait_info_function_table[trait_type])
 
-    return retrieve_info(
+    trait_dataset = retrieve_trait_dataset(trait_type, trait, threshold, conn)
+    trait_info = retrieve_info(
         {
             "trait_name": trait["trait_name"],
-            "trait_dataset_id": trait_dataset_id,
-            "trait_dataset_name":trait_dataset_name
+            "trait_dataset_id": trait_dataset["dataset_id"],
+            "trait_dataset_name": trait_dataset["dataset_name"]
         },
         conn)
+    return {
+        **trait_info,
+        "db": {**trait["db"], **trait_dataset},
+        "riset": trait_dataset["riset"]
+    }