From be8925e163b9829daadb645a83caedc933f98978 Mon Sep 17 00:00:00 2001
From: Emmanuel Viennet <emmanuel.viennet@gmail.com>
Date: Sun, 28 Nov 2021 16:27:05 +0100
Subject: [PATCH] charegment matrice inscriptions modules

---
 app/comp/inscr_mod.py | 70 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 70 insertions(+)
 create mode 100644 app/comp/inscr_mod.py

diff --git a/app/comp/inscr_mod.py b/app/comp/inscr_mod.py
new file mode 100644
index 000000000..d3dad9b43
--- /dev/null
+++ b/app/comp/inscr_mod.py
@@ -0,0 +1,70 @@
+# -*- mode: python -*-
+# -*- coding: utf-8 -*-
+
+"""Matrices d'inscription aux modules d'un semestre
+"""
+import numpy as np
+import pandas as pd
+
+from app import db
+from app import models
+
+#
+# Le chargement des inscriptions est long: matrice nb_module x nb_etuds
+# sur test debug 116 etuds, 18 modules, on est autour de 250ms.
+# On a testé trois approches, ci-dessous (et retenu la 1ere)
+#
+def df_load_modimpl_inscr(formsemestre_id):
+    """Charge la matrice des inscriptions aux modules du semestre
+    rows: etudid
+    columns: moduleimpl_id
+    value: int (0/1 inscrit ou pas)
+    """
+    # méthode la moins lente: une requete par module, merge les dataframes
+    sem = models.FormSemestre.query.get(formsemestre_id)
+    moduleimpl_ids = [m.id for m in sem.modimpls]
+    etudids = [i.etudid for i in sem.inscriptions]
+    df = pd.DataFrame(index=etudids, dtype=int)
+    for moduleimpl_id in moduleimpl_ids:
+        ins_df = pd.read_sql_query(
+            """SELECT etudid, 1 AS "%(moduleimpl_id)s"
+                FROM notes_moduleimpl_inscription
+                WHERE moduleimpl_id=%(moduleimpl_id)s""",
+            db.engine,
+            params={"moduleimpl_id": moduleimpl_id},
+            index_col="etudid",
+            dtype=int,
+        )
+        df = df.merge(ins_df, how="outer", left_index=True, right_index=True)
+    return df  # x100 25.5s 15s 17s
+
+
+# chrono avec timeit:
+# timeit.timeit('x = df_load_module_inscr_v0(696)', number=100, globals=globals())
+
+
+def df_load_modimpl_inscr_v0(formsemestre_id):
+    # methode 0, pur SQL Alchemy, 1.5 à 2 fois plus lente
+    sem = models.FormSemestre.query.get(formsemestre_id)
+    moduleimpl_ids = [m.id for m in sem.modimpls]
+    etudids = [i.etudid for i in sem.inscriptions]
+    df = pd.DataFrame(False, columns=moduleimpl_ids, index=etudids, dtype=bool)
+    for modimpl in sem.modimpls:
+        ins_mod = df[modimpl.id]
+        for inscr in modimpl.inscriptions:
+            ins_mod[inscr.etudid] = True
+    return df  # x100 30.7s 46s 32s
+
+
+def df_load_modimpl_inscr_v2(formsemestre_id):
+    sem = models.FormSemestre.query.get(formsemestre_id)
+    moduleimpl_ids = [m.id for m in sem.modimpls]
+    etudids = [i.etudid for i in sem.inscriptions]
+    df = pd.DataFrame(False, columns=moduleimpl_ids, index=etudids, dtype=bool)
+    cursor = db.engine.execute(
+        "select moduleimpl_id, etudid from notes_moduleimpl_inscription i, notes_moduleimpl m where i.moduleimpl_id = m.id and m.formsemestre_id = %(formsemestre_id)s",
+        {"formsemestre_id": formsemestre_id},
+    )
+    for moduleimpl_id, etudid in cursor:
+        df[moduleimpl_id][etudid] = True
+    return df  # x100 44s, 31s, 29s, 28s