forked from ScoDoc/DocScoDoc
374 lines
13 KiB
Python
374 lines
13 KiB
Python
# -*- mode: python -*-
|
|
# -*- coding: utf-8 -*-
|
|
|
|
##############################################################################
|
|
#
|
|
# Gestion scolarite IUT
|
|
#
|
|
# Copyright (c) 1999 - 2023 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
|
|
#
|
|
##############################################################################
|
|
|
|
"""Comparaison de deux fichiers Apogée (maquettes)
|
|
|
|
1) Vérifier:
|
|
etape_apogee, vdi_apogee, cod_dip_apogee, annee_scolaire
|
|
structure: col_ids (la comparaison portera sur les colonnes communes)
|
|
|
|
|
|
2) Comparer listes d'étudiants
|
|
Présents dans A mais pas dans B
|
|
Présents dans B mais pas dans A
|
|
nombre communs
|
|
|
|
3) Comparer résultats
|
|
Pour chaque étudiant commun:
|
|
Pour chaque colonne commune:
|
|
comparer les résultats
|
|
|
|
"""
|
|
from flask import g, url_for
|
|
|
|
from app import log
|
|
from app.scodoc import sco_apogee_csv
|
|
from app.scodoc.gen_tables import GenTable
|
|
from app.scodoc.sco_exceptions import ScoValueError
|
|
from app.scodoc import html_sco_header
|
|
from app.scodoc import sco_preferences
|
|
|
|
_help_txt = """
|
|
<div class="help">
|
|
<p>Outil de comparaison de fichiers (maquettes CSV) Apogée.
|
|
</p>
|
|
<p>Cet outil compare deux fichiers fournis. Aucune donnée stockée dans ScoDoc n'est utilisée.
|
|
</p>
|
|
</div>
|
|
"""
|
|
|
|
|
|
def apo_compare_csv_form():
|
|
"""Form: submit 2 CSV files to compare them."""
|
|
H = [
|
|
html_sco_header.sco_header(page_title="Comparaison de fichiers Apogée"),
|
|
"""<h2>Comparaison de fichiers Apogée</h2>
|
|
<form id="apo_csv_add" action="apo_compare_csv" method="post" enctype="multipart/form-data">
|
|
""",
|
|
_help_txt,
|
|
"""
|
|
<div class="apo_compare_csv_form_but">
|
|
Fichier Apogée A:
|
|
<input type="file" size="30" name="file_a"/>
|
|
</div>
|
|
<div class="apo_compare_csv_form_but">
|
|
Fichier Apogée B:
|
|
<input type="file" size="30" name="file_b"/>
|
|
</div>
|
|
<input type="checkbox" name="autodetect" checked/>autodétecter encodage</input>
|
|
<div class="apo_compare_csv_form_submit">
|
|
<input type="submit" value="Comparer ces fichiers"/>
|
|
</div>
|
|
</form>""",
|
|
html_sco_header.sco_footer(),
|
|
]
|
|
return "\n".join(H)
|
|
|
|
|
|
def apo_compare_csv(file_a, file_b, autodetect=True):
|
|
"""Page comparing 2 Apogee CSV files"""
|
|
try:
|
|
apo_data_a = _load_apo_data(file_a, autodetect=autodetect)
|
|
apo_data_b = _load_apo_data(file_b, autodetect=autodetect)
|
|
except (UnicodeDecodeError, UnicodeEncodeError) as exc:
|
|
dest_url = url_for("notes.semset_page", scodoc_dept=g.scodoc_dept)
|
|
if autodetect:
|
|
raise ScoValueError(
|
|
"""
|
|
Erreur: l'encodage de l'un des fichiers est mal détecté.
|
|
Essayez sans auto-détection, ou vérifiez le codage et le contenu
|
|
des fichiers.
|
|
""",
|
|
dest_url=dest_url,
|
|
) from exc
|
|
else:
|
|
raise ScoValueError(
|
|
f"""
|
|
Erreur: l'encodage de l'un des fichiers est incorrect.
|
|
Vérifiez qu'il est bien en {sco_apogee_csv.APO_INPUT_ENCODING}
|
|
""",
|
|
dest_url=dest_url,
|
|
) from exc
|
|
H = [
|
|
html_sco_header.sco_header(page_title="Comparaison de fichiers Apogée"),
|
|
"<h2>Comparaison de fichiers Apogée</h2>",
|
|
_help_txt,
|
|
'<div class="apo_compare_csv">',
|
|
_apo_compare_csv(apo_data_a, apo_data_b),
|
|
"</div>",
|
|
"""<p><a href="apo_compare_csv_form" class="stdlink">Autre comparaison</a></p>""",
|
|
html_sco_header.sco_footer(),
|
|
]
|
|
return "\n".join(H)
|
|
|
|
|
|
def _load_apo_data(csvfile, autodetect=True):
|
|
"Read data from request variable and build ApoData"
|
|
data_b = csvfile.read()
|
|
if autodetect:
|
|
data_b, message = sco_apogee_csv.fix_data_encoding(data_b)
|
|
if message:
|
|
log(f"apo_compare_csv: {message}")
|
|
if not data_b:
|
|
raise ScoValueError("fichier vide ? (apo_compare_csv: no data)")
|
|
data = data_b.decode(sco_apogee_csv.APO_INPUT_ENCODING)
|
|
apo_data = sco_apogee_csv.ApoData(data, orig_filename=csvfile.filename)
|
|
return apo_data
|
|
|
|
|
|
def _apo_compare_csv(A, B):
|
|
"""Generate html report comparing A and B, two instances of ApoData
|
|
representing Apogee CSV maquettes.
|
|
"""
|
|
L = []
|
|
# 1-- Check etape and codes
|
|
L.append('<div class="section"><div class="tit">En-tête</div>')
|
|
L.append('<div><span class="key">Nom fichier A:</span><span class="val_ok">')
|
|
L.append(A.orig_filename)
|
|
L.append("</span></div>")
|
|
L.append('<div><span class="key">Nom fichier B:</span><span class="val_ok">')
|
|
L.append(B.orig_filename)
|
|
L.append("</span></div>")
|
|
L.append('<div><span class="key">Étape Apogée:</span>')
|
|
if A.etape_apogee != B.etape_apogee:
|
|
L.append(
|
|
'<span class="val_dif">%s != %s</span>' % (A.etape_apogee, B.etape_apogee)
|
|
)
|
|
else:
|
|
L.append('<span class="val_ok">%s</span>' % (A.etape_apogee,))
|
|
L.append("</div>")
|
|
|
|
L.append('<div><span class="key">VDI Apogée:</span>')
|
|
if A.vdi_apogee != B.vdi_apogee:
|
|
L.append('<span class="val_dif">%s != %s</span>' % (A.vdi_apogee, B.vdi_apogee))
|
|
else:
|
|
L.append('<span class="val_ok">%s</span>' % (A.vdi_apogee,))
|
|
L.append("</div>")
|
|
|
|
L.append('<div><span class="key">Code diplôme :</span>')
|
|
if A.cod_dip_apogee != B.cod_dip_apogee:
|
|
L.append(
|
|
'<span class="val_dif">%s != %s</span>'
|
|
% (A.cod_dip_apogee, B.cod_dip_apogee)
|
|
)
|
|
else:
|
|
L.append('<span class="val_ok">%s</span>' % (A.cod_dip_apogee,))
|
|
L.append("</div>")
|
|
|
|
L.append('<div><span class="key">Année scolaire :</span>')
|
|
if A.annee_scolaire != B.annee_scolaire:
|
|
L.append(
|
|
'<span class="val_dif">%s != %s</span>'
|
|
% (A.annee_scolaire, B.annee_scolaire)
|
|
)
|
|
else:
|
|
L.append('<span class="val_ok">%s</span>' % (A.annee_scolaire,))
|
|
L.append("</div>")
|
|
|
|
# Colonnes:
|
|
A_elts = set(A.apo_elts.keys())
|
|
B_elts = set(B.apo_elts.keys())
|
|
L.append('<div><span class="key">Éléments Apogée :</span>')
|
|
if A_elts == B_elts:
|
|
L.append('<span class="val_ok">%d</span>' % len(A_elts))
|
|
else:
|
|
elts_communs = A_elts.intersection(B_elts)
|
|
elts_only_A = A_elts - A_elts.intersection(B_elts)
|
|
elts_only_B = B_elts - A_elts.intersection(B_elts)
|
|
L.append(
|
|
'<span class="val_dif">différents (%d en commun, %d seulement dans A, %d seulement dans B)</span>'
|
|
% (
|
|
len(elts_communs),
|
|
len(elts_only_A),
|
|
len(elts_only_B),
|
|
)
|
|
)
|
|
if elts_only_A:
|
|
L.append(
|
|
'<div span class="key">Éléments seulement dans A : </span><span class="val_dif">%s</span></div>'
|
|
% ", ".join(sorted(elts_only_A))
|
|
)
|
|
if elts_only_B:
|
|
L.append(
|
|
'<div span class="key">Éléments seulement dans B : </span><span class="val_dif">%s</span></div>'
|
|
% ", ".join(sorted(elts_only_B))
|
|
)
|
|
L.append("</div>")
|
|
L.append("</div>") # /section
|
|
|
|
# 2--
|
|
L.append('<div class="section"><div class="tit">Étudiants</div>')
|
|
|
|
A_nips = set(A.etud_by_nip)
|
|
B_nips = set(B.etud_by_nip)
|
|
nb_etuds_communs = len(A_nips.intersection(B_nips))
|
|
nb_etuds_dif = len(A_nips.union(B_nips) - A_nips.intersection(B_nips))
|
|
L.append("""<div><span class="key">Liste d'étudiants :</span>""")
|
|
if A_nips == B_nips:
|
|
L.append(
|
|
"""<span class="s_ok">
|
|
%d étudiants (tous présents dans chaque fichier)</span>
|
|
"""
|
|
% len(A_nips)
|
|
)
|
|
else:
|
|
L.append(
|
|
'<span class="val_dif">différents (%d en commun, %d différents)</span>'
|
|
% (nb_etuds_communs, nb_etuds_dif)
|
|
)
|
|
L.append("</div>")
|
|
L.append("</div>") # /section
|
|
|
|
# 3-- Résultats de chaque étudiant:
|
|
if nb_etuds_communs > 0:
|
|
L.append(
|
|
"""<div class="section sec_table">
|
|
<div class="tit">Différences de résultats des étudiants présents dans les deux fichiers</div>
|
|
<p>
|
|
"""
|
|
)
|
|
T = apo_table_compare_etud_results(A, B)
|
|
if T.get_nb_rows() > 0:
|
|
L.append(T.html())
|
|
else:
|
|
L.append(
|
|
"""<p class="p_ok">aucune différence de résultats
|
|
sur les %d étudiants communs (<em>les éléments Apogée n'apparaissant pas dans les deux fichiers sont omis</em>)</p>
|
|
"""
|
|
% nb_etuds_communs
|
|
)
|
|
L.append("</div>") # /section
|
|
|
|
return "\n".join(L)
|
|
|
|
|
|
def apo_table_compare_etud_results(A, B):
|
|
""""""
|
|
D = compare_etuds_res(A, B)
|
|
T = GenTable(
|
|
rows=D,
|
|
titles={
|
|
"nip": "NIP",
|
|
"nom": "Nom",
|
|
"prenom": "Prénom",
|
|
"elt_code": "Element",
|
|
"type_res": "Type",
|
|
"val_A": "A: %s" % (A.orig_filename or ""),
|
|
"val_B": "B: %s" % (B.orig_filename or ""),
|
|
},
|
|
columns_ids=("nip", "nom", "prenom", "elt_code", "type_res", "val_A", "val_B"),
|
|
html_class="table_leftalign",
|
|
html_with_td_classes=True,
|
|
preferences=sco_preferences.SemPreferences(),
|
|
)
|
|
return T
|
|
|
|
|
|
def _build_etud_res(e, apo_data):
|
|
r = {}
|
|
for elt_code in apo_data.apo_elts:
|
|
elt = apo_data.apo_elts[elt_code]
|
|
try:
|
|
# les colonnes de cet élément
|
|
col_ids_type = [
|
|
(ec["apoL_a01_code"], ec["Type R\xc3\xa9s."]) for ec in elt.cols
|
|
]
|
|
except KeyError as exc:
|
|
raise ScoValueError(
|
|
"Erreur: un élément sans 'Type R\xc3\xa9s.'. Vérifiez l'encodage de vos fichiers."
|
|
) from exc
|
|
r[elt_code] = {}
|
|
for (col_id, type_res) in col_ids_type:
|
|
r[elt_code][type_res] = e.cols[col_id]
|
|
return r
|
|
|
|
|
|
def compare_etud_res(r_A, r_B, remove_missing=True):
|
|
"""Pour chaque valeur difference dans les resultats d'un etudiant
|
|
elt_code type_res val_A val_B
|
|
"""
|
|
diffs = []
|
|
elt_codes = set(r_A).union(set(r_B))
|
|
for elt_code in elt_codes:
|
|
for type_res in r_A.get(elt_code, r_B.get(elt_code)):
|
|
if elt_code not in r_A:
|
|
if remove_missing:
|
|
continue
|
|
else:
|
|
val_A = None # element absent
|
|
else:
|
|
val_A = r_A[elt_code][type_res]
|
|
if elt_code not in r_B:
|
|
if remove_missing:
|
|
continue
|
|
else:
|
|
val_B = None # element absent
|
|
else:
|
|
val_B = r_B[elt_code][type_res]
|
|
if type_res == "N":
|
|
# Cas particulier pour les notes: compare les nombres
|
|
try:
|
|
val_A_num = float(val_A.replace(",", "."))
|
|
val_B_num = float(val_B.replace(",", "."))
|
|
except ValueError:
|
|
val_A_num, val_B_num = val_A, val_B
|
|
val_A, val_B = val_A_num, val_B_num
|
|
if val_A != val_B:
|
|
diffs.append(
|
|
{
|
|
"elt_code": elt_code,
|
|
"type_res": type_res,
|
|
"val_A": val_A,
|
|
"val_B": val_B,
|
|
}
|
|
)
|
|
return diffs
|
|
|
|
|
|
def compare_etuds_res(A, B):
|
|
"""
|
|
nip, nom, prenom, elt_code, type_res, val_A, val_B
|
|
"""
|
|
A_nips = set(A.etud_by_nip)
|
|
B_nips = set(B.etud_by_nip)
|
|
common_nips = A_nips.intersection(B_nips)
|
|
# A_not_B_nips = A_nips - B_nips
|
|
# B_not_A_nips = B_nips - A_nips
|
|
D = []
|
|
for nip in common_nips:
|
|
etu_A = A.etud_by_nip[nip]
|
|
etu_B = B.etud_by_nip[nip]
|
|
r_A = _build_etud_res(etu_A, A)
|
|
r_B = _build_etud_res(etu_B, B)
|
|
diffs = compare_etud_res(r_A, r_B)
|
|
for d in diffs:
|
|
d.update(
|
|
{"nip": etu_A["nip"], "nom": etu_A["nom"], "prenom": etu_A["prenom"]}
|
|
)
|
|
D.append(d)
|
|
return D
|