217 lines
8.3 KiB
Python
217 lines
8.3 KiB
Python
# -*- mode: python -*-
|
|
# -*- coding: utf-8 -*-
|
|
|
|
##############################################################################
|
|
#
|
|
# Gestion scolarite IUT
|
|
#
|
|
# Copyright (c) 1999 - 2021 Emmanuel Viennet. All rights reserved.
|
|
#
|
|
# This program is free software; you can redistribute it and/or modify
|
|
# it under the terms of the GNU General Public License as published by
|
|
# the Free Software Foundation; either version 2 of the License, or
|
|
# (at your option) any later version.
|
|
#
|
|
# This program is distributed in the hope that it will be useful,
|
|
# but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
# GNU General Public License for more details.
|
|
#
|
|
# You should have received a copy of the GNU General Public License
|
|
# along with this program; if not, write to the Free Software
|
|
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
|
#
|
|
# Emmanuel Viennet emmanuel.viennet@viennet.net
|
|
#
|
|
##############################################################################
|
|
|
|
"""Fonctions de calcul des moyennes d'UE
|
|
"""
|
|
import numpy as np
|
|
import pandas as pd
|
|
|
|
from app import db
|
|
from app import models
|
|
from app.models import UniteEns, Module, ModuleImpl, ModuleUECoef
|
|
from app.comp import moy_mod
|
|
from app.models.formsemestre import FormSemestre
|
|
from app.scodoc import sco_codes_parcours
|
|
|
|
|
|
def df_load_module_coefs(formation_id: int, semestre_idx: int = None) -> pd.DataFrame:
|
|
"""Charge les coefs des modules de la formation pour le semestre indiqué.
|
|
|
|
Ces coefs lient les modules à chaque UE.
|
|
|
|
Résultat: (module_coefs_df, ues, modules)
|
|
DataFrame rows = UEs, columns = modules, value = coef.
|
|
|
|
Considère toutes les UE (sauf sport) et modules du semestre.
|
|
Les coefs non définis (pas en base) sont mis à zéro.
|
|
|
|
Si semestre_idx None, prend toutes les UE de la formation.
|
|
"""
|
|
ues = (
|
|
UniteEns.query.filter_by(formation_id=formation_id)
|
|
.filter(UniteEns.type != sco_codes_parcours.UE_SPORT)
|
|
.order_by(UniteEns.semestre_idx, UniteEns.numero, UniteEns.acronyme)
|
|
)
|
|
modules = Module.query.filter_by(formation_id=formation_id).order_by(
|
|
Module.semestre_id, Module.module_type.desc(), Module.numero, Module.code
|
|
)
|
|
if semestre_idx is not None:
|
|
ues = ues.filter_by(semestre_idx=semestre_idx)
|
|
modules = modules.filter_by(semestre_id=semestre_idx)
|
|
ues = ues.all()
|
|
modules = modules.all()
|
|
ue_ids = [ue.id for ue in ues]
|
|
module_ids = [module.id for module in modules]
|
|
module_coefs_df = pd.DataFrame(columns=module_ids, index=ue_ids, dtype=float)
|
|
for mod_coef in (
|
|
db.session.query(ModuleUECoef)
|
|
.filter(UniteEns.formation_id == formation_id)
|
|
.filter(ModuleUECoef.ue_id == UniteEns.id)
|
|
):
|
|
module_coefs_df[mod_coef.module_id][mod_coef.ue_id] = mod_coef.coef
|
|
module_coefs_df.fillna(value=0, inplace=True)
|
|
return module_coefs_df, ues, modules
|
|
|
|
|
|
def df_load_modimpl_coefs(
|
|
formsemestre: models.FormSemestre, ues=None, modimpls=None
|
|
) -> pd.DataFrame:
|
|
"""Charge les coefs des modules du formsemestre indiqué.
|
|
|
|
Comme df_load_module_coefs mais prend seulement les UE
|
|
et modules du formsemestre.
|
|
Si ues et modimpls sont None, prend tous ceux du formsemestre.
|
|
Résultat: (module_coefs_df, ues, modules)
|
|
DataFrame rows = UEs, columns = modimpl, value = coef.
|
|
"""
|
|
if ues is None:
|
|
ues = formsemestre.query_ues().all()
|
|
ue_ids = [x.id for x in ues]
|
|
if modimpls is None:
|
|
modimpls = formsemestre.modimpls.all()
|
|
modimpl_ids = [x.id for x in modimpls]
|
|
mod2impl = {m.module.id: m.id for m in modimpls}
|
|
modimpl_coefs_df = pd.DataFrame(columns=modimpl_ids, index=ue_ids, dtype=float)
|
|
mod_coefs = (
|
|
db.session.query(ModuleUECoef)
|
|
.filter(ModuleUECoef.module_id == ModuleImpl.module_id)
|
|
.filter(ModuleImpl.formsemestre_id == formsemestre.id)
|
|
)
|
|
|
|
for mod_coef in mod_coefs:
|
|
modimpl_coefs_df[mod2impl[mod_coef.module_id]][mod_coef.ue_id] = mod_coef.coef
|
|
modimpl_coefs_df.fillna(value=0, inplace=True)
|
|
return modimpl_coefs_df, ues, modimpls
|
|
|
|
|
|
def notes_sem_assemble_cube(modimpls_notes: list[pd.DataFrame]) -> np.ndarray:
|
|
"""Réuni les notes moyennes des modules du semestre en un "cube"
|
|
|
|
modimpls_notes : liste des moyennes de module
|
|
(DataFrames rendus par compute_module_moy, (etud x UE))
|
|
Resultat: ndarray (etud x module x UE)
|
|
"""
|
|
modimpls_notes_arr = [df.values for df in modimpls_notes]
|
|
modimpls_notes = np.stack(modimpls_notes_arr)
|
|
# passe de (mod x etud x ue) à (etud x mod x UE)
|
|
return modimpls_notes.swapaxes(0, 1)
|
|
|
|
|
|
def notes_sem_load_cube(formsemestre):
|
|
"""Calcule le cube des notes du semestre
|
|
(charge toutes les notes, calcule les moyenne des modules
|
|
et assemble le cube)
|
|
Resultat:
|
|
sem_cube : ndarray (etuds x modimpls x UEs)
|
|
modimpls_evals_poids dict { modimpl.id : evals_poids }
|
|
modimpls_evals_notes dict { modimpl.id : evals_notes }
|
|
modimpls_evaluations dict { modimpl.id : liste des évaluations }
|
|
"""
|
|
modimpls_evals_poids = {}
|
|
modimpls_evals_notes = {}
|
|
modimpls_evaluations = {}
|
|
modimpls_notes = []
|
|
for modimpl in formsemestre.modimpls:
|
|
evals_notes, evaluations, evaluations_completes = moy_mod.df_load_modimpl_notes(
|
|
modimpl.id
|
|
)
|
|
evals_poids, ues = moy_mod.df_load_evaluations_poids(modimpl.id)
|
|
etuds_moy_module = moy_mod.compute_module_moy(
|
|
evals_notes, evals_poids, evaluations, evaluations_completes
|
|
)
|
|
modimpls_evals_poids[modimpl.id] = evals_poids
|
|
modimpls_evals_notes[modimpl.id] = evals_notes
|
|
modimpls_evaluations[modimpl.id] = evaluations
|
|
modimpls_notes.append(etuds_moy_module)
|
|
return (
|
|
notes_sem_assemble_cube(modimpls_notes),
|
|
modimpls_evals_poids,
|
|
modimpls_evals_notes,
|
|
modimpls_evaluations,
|
|
)
|
|
|
|
|
|
def compute_ue_moys(
|
|
sem_cube: np.array,
|
|
etuds: list,
|
|
modimpls: list,
|
|
ues: list,
|
|
modimpl_inscr_df: pd.DataFrame,
|
|
modimpl_coefs_df: pd.DataFrame,
|
|
) -> pd.DataFrame:
|
|
"""Calcul de la moyenne d'UE
|
|
La moyenne d'UE est un nombre (note/20), ou NI ou NA ou ERR
|
|
NI non inscrit à (au moins un) module de cette UE
|
|
NA pas de notes disponibles
|
|
ERR erreur dans une formule utilisateur. [XXX pas encore gérées ici]
|
|
|
|
sem_cube: notes moyennes aux modules
|
|
ndarray (etuds x modimpls x UEs)
|
|
(floats avec des NaN)
|
|
etuds : lites des étudiants (dim. 0 du cube)
|
|
modimpls : liste des modules à considérer (dim. 1 du cube)
|
|
ues : liste des UE (dim. 2 du cube)
|
|
module_inscr_df: matrice d'inscription du semestre (etud x modimpl)
|
|
module_coefs_df: matrice coefficients (UE x modimpl)
|
|
|
|
Resultat: DataFrame columns UE, rows etudid
|
|
"""
|
|
nb_etuds, nb_modules, nb_ues = sem_cube.shape
|
|
assert len(etuds) == nb_etuds
|
|
assert len(modimpls) == nb_modules
|
|
assert len(ues) == nb_ues
|
|
assert modimpl_inscr_df.shape[0] == nb_etuds
|
|
assert modimpl_inscr_df.shape[1] == nb_modules
|
|
assert modimpl_coefs_df.shape[0] == nb_ues
|
|
assert modimpl_coefs_df.shape[1] == nb_modules
|
|
modimpl_inscr = modimpl_inscr_df.values
|
|
modimpl_coefs = modimpl_coefs_df.values
|
|
# Duplique les inscriptions sur les UEs:
|
|
modimpl_inscr_stacked = np.stack([modimpl_inscr] * nb_ues, axis=2)
|
|
# Enlève les NaN du numérateur:
|
|
# si on veut prendre en compte les modules avec notes neutralisées ?
|
|
sem_cube_no_nan = np.nan_to_num(sem_cube, nan=0.0)
|
|
|
|
# Ne prend pas en compte les notes des étudiants non inscrits au module:
|
|
# Annule les notes:
|
|
sem_cube_inscrits = np.where(modimpl_inscr_stacked, sem_cube_no_nan, 0.0)
|
|
# Annule les coefs des modules où l'étudiant n'est pas inscrit:
|
|
modimpl_coefs_etuds = np.where(
|
|
modimpl_inscr_stacked, np.stack([modimpl_coefs.T] * nb_etuds), 0.0
|
|
)
|
|
# Annule les coefs des modules NaN
|
|
modimpl_coefs_etuds_no_nan = np.where(np.isnan(sem_cube), 0.0, modimpl_coefs_etuds)
|
|
#
|
|
# Version vectorisée
|
|
#
|
|
etud_moy_ue = np.sum(
|
|
modimpl_coefs_etuds_no_nan * sem_cube_inscrits, axis=1
|
|
) / np.sum(modimpl_coefs_etuds_no_nan, axis=1)
|
|
return pd.DataFrame(
|
|
etud_moy_ue, index=modimpl_inscr_df.index, columns=modimpl_coefs_df.index
|
|
)
|