ScoDoc-PE/app/comp/inscr_mod.py

# -*- 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) -> pd.DataFrame:
    """Charge la matrice des inscriptions aux modules du semestre
    rows: etudid
    columns: moduleimpl_id (en chaîne)
    value: bool (0/1 inscrit ou pas)
    """
    # méthode la moins lente: une requete par module, merge les dataframes
    moduleimpl_ids = [m.id for m in formsemestre.modimpls]
    etudids = [i.etudid for i in formsemestre.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)
    # les colonnes de df sont en float (Nan) quand il n'y a
    # aucun inscrit au module.
    df.fillna(0, inplace=True)  # les non-inscrits
    return df.astype(bool)  # 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):
    # methode 0, pur SQL Alchemy, 1.5 à 2 fois plus lente
    moduleimpl_ids = [m.id for m in formsemestre.modimpls]
    etudids = [i.etudid for i in formsemestre.inscriptions]
    df = pd.DataFrame(False, columns=moduleimpl_ids, index=etudids, dtype=bool)
    for modimpl in formsemestre.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):
    moduleimpl_ids = [m.id for m in formsemestre.modimpls]
    etudids = [i.etudid for i in formsemestre.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