File size: 10,408 Bytes
f0806e2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 |
import time
import pandas as pd
from collections import namedtuple
from tqdm import tqdm
import ast
from concurrent.futures import ThreadPoolExecutor
import os
import multiprocessing
from neo4j.exceptions import TransientError
# --- Paramètres ---
BATCH_SIZE = 5000
NUM_THREADS = multiprocessing.cpu_count()
PROCESSED_IDS_FILE = "processed_ids.txt"
def reset_database(driver):
"""
Efface TOUTES les données de la base Neo4j ET le fichier de suivi des IDs traités.
À n'utiliser que pour une réinitialisation complète.
"""
# Étape 1 : Vider la base de données
with driver.session() as session:
result = session.run("RETURN 1 AS test")
print("Connexion OK, test result:", result.single()["test"])
session.run("MATCH (n) DETACH DELETE n")
# Étape 2 : Supprimer le fichier de suivi pour garantir une réimportation propre
if os.path.exists(PROCESSED_IDS_FILE):
os.remove(PROCESSED_IDS_FILE)
def parse_list_field(value):
"""
Parse une chaîne de caractères qui représente une liste (ex: "['model1', 'model2']")
en une véritable liste Python. Gère les cas où la valeur est simple ou vide.
"""
if isinstance(value, str) and pd.notna(value):
try:
parsed = ast.literal_eval(value)
if isinstance(parsed, list):
return parsed
except Exception:
pass
return [value] if value else []
def load_processed_ids():
"""Charge l'ensemble des IDs déjà traités depuis le fichier de suivi."""
if os.path.exists(PROCESSED_IDS_FILE):
with open(PROCESSED_IDS_FILE, "r", encoding="utf-8") as f:
return set(line.strip() for line in f)
return set()
def append_processed_ids(ids):
"""Ajoute une liste d'IDs au fichier de suivi."""
with open(PROCESSED_IDS_FILE, "a", encoding="utf-8") as f:
for i in ids:
f.write(f"{i}\n")
def run_with_retry(session, query, parameters=None, retries=3, delay=1):
"""
Exécute une requête Cypher avec une logique de réessai en cas d'erreur transitoire
"""
for attempt in range(retries):
try:
session.run(query, parameters)
return
except TransientError as e:
if attempt < retries - 1:
time.sleep(delay)
continue
else:
raise
def process_batch(rows, fieldnames, driver):
"""
Traite un lot (batch) de lignes du CSV et les insère dans Neo4j.
C'est la fonction "worker" qui sera exécutée en parallèle.
"""
normalized_fields = [f.strip().replace(" ", "_").replace("-", "_") for f in fieldnames]
Row = namedtuple("Row", normalized_fields)
ids_successfully_processed = []
with driver.session() as session:
for row in rows:
obj = Row(**{k: v if v != "" else None for k, v in row.items()})
data = obj._asdict()
if not data.get("id") or pd.isna(data.get("id")) or pd.isna(data.get("author")) or str(data.get("id")).strip() == "":
continue # Ignore si l'ID est manquant
base_models = parse_list_field(data.get("base_model"))
base_model_rels = parse_list_field(data.get("base_model_relation"))
datasets = parse_list_field(data.get("dataset"))
orgs_author_model = parse_list_field(data.get("organizations_author_model"))
orgs_author_dataset = parse_list_field(data.get("organizations_author_dataset"))
# Insertion du modèle et de son auteur
if data.get("id") and data.get("author"):
run_with_retry(session, """
MERGE (m:Model {name: $id})
SET m.downloads = $downloadsAllTime,
m.task = $pipeline_tag,
m.createdAt = $createdAt,
m.parameters = $total_parameters_formatted,
m.likes = $likes,
m.license = $license
MERGE (a:Author {name: $author})
SET a.type = $author_type,
a.followers = $followers_count_author_model
WITH a
MATCH (m:Model {name: $id})
MERGE (a)-[p:POSTED]->(m)
SET p.name = "A publié"
""", data)
# Lien entre l’auteur et ses organisations
orgs_data = [
{"org": org, "author": data["author"]}
for org in orgs_author_model
if pd.notna(org) and data.get("author")
]
if orgs_data:
run_with_retry(session, """
UNWIND $orgs_data AS row
MERGE (o:Author {name: row.org})
WITH o, row
MATCH (a:Author {name: row.author})
MERGE (a)-[r:IS_IN]->(o)
SET r.name = "Fait partie de cette organisation", a.type = "personne",o.type = "organisation"
""", {"orgs_data": orgs_data})
# Lien entre modèles et base models
if len(base_models) == len(base_model_rels) :
base_model_data = [
{"bm": bm, "id": data["id"], "rel": rel}
for bm, rel in zip(base_models, base_model_rels)
if pd.notna(bm) and data.get("id")
]
elif len(base_models) >len(base_model_rels) :
if base_model_rels==['merge'] :
base_model_data = [
{"bm": bm, "id": data["id"], "rel": "merge"}
for bm in base_models
if pd.notna(bm) and data.get("id")
]
else :
base_model_data = [
{"bm": bm, "id": data["id"], "rel": "A généré"}
for bm in base_models
if pd.notna(bm) and data.get("id")
]
if base_model_data:
run_with_retry(session, """
UNWIND $base_model_data AS row
MERGE (bm:Model {name: row.bm})
WITH bm, row
MATCH (m:Model {name: row.id})
MERGE (bm)-[r:USED_IN]->(m)
SET r.name = row.rel
""", {"base_model_data": base_model_data})
# Lien entre modèles et datasets
datasets_data = [
{"ds": ds, "downloads": data.get("downloads_dataset"),
"createdAt_dataset": data.get("createdAt_dataset"), "id": data["id"]}
for ds in datasets
if pd.notna(ds) and data.get("id")
]
if datasets_data and data.get("author_dataset") and data.get("dataset") and pd.notna(data.get("author_dataset")):
run_with_retry(session, """
UNWIND $datasets_data AS row
MERGE (d:Dataset {name: row.ds})
SET d.downloads = row.downloads,
d.createdAt_dataset = row.createdAt_dataset
WITH d, row
MATCH (m:Model {name: row.id})
MERGE (d)-[r:USED_IN]->(m)
SET r.name = "A été utilisé dans ce modèle"
""", {"datasets_data": datasets_data})
# Insertion de l’auteur du dataset
run_with_retry(session, """
MERGE (ad:Author {name: $author_dataset})
SET ad.type = $author_dataset_type,
ad.followers = $followers_count_author_dataset
WITH ad
MATCH (d:Dataset {name: $dataset})
MERGE (ad)-[r:POSTED]->(d)
SET r.name = "A publié"
""", data)
# Lien entre l’auteur du dataset et ses organisations
orgs_dataset_data = [
{"org": org, "author_dataset": data["author_dataset"]}
for org in orgs_author_dataset
if pd.notna(org) and data.get("author_dataset")
]
if orgs_dataset_data and pd.notna(data.get("author_dataset")):
run_with_retry(session, """
UNWIND $orgs_data AS row
MERGE (o:Author {name: row.org})
WITH o, row
MATCH (ad:Author {name: row.author_dataset})
MERGE (ad)-[r:IS_IN]->(o)
SET r.name = "Fait partie de cette organisation", ad.type = "personne",o.type = "organisation"
""", {"orgs_data": orgs_dataset_data})
ids_successfully_processed.append(data["id"])
if ids_successfully_processed:
append_processed_ids(ids_successfully_processed)
# Insère les données depuis un CSV en parallèle, par lots
def insert_parallel(csv_file_path, driver, processed_ids):
# Lecture et nettoyage via pandas
df = pd.read_csv(csv_file_path)
df = df.loc[:, ~df.columns.str.contains('^Unnamed')]
# Supprimer les lignes où 'id' est NaN ou vide
df = df[~df["id"].isnull()]
df = df[df["id"].astype(str).str.strip() != ""]
# Ne conserver que les lignes dont l'ID n’a pas encore été traitée
df = df[~df["id"].isin(processed_ids)]
records = df.to_dict(orient="records")
fieldnames = list(df.columns)
batch = []
futures = []
with ThreadPoolExecutor(max_workers=NUM_THREADS) as executor:
for row in tqdm(records, desc="Lecture CSV"):
batch.append(row)
if len(batch) == BATCH_SIZE:
futures.append(executor.submit(process_batch, batch.copy(), fieldnames, driver))
batch = []
if batch:
futures.append(executor.submit(process_batch, batch.copy(), fieldnames, driver))
for future in tqdm(futures, desc="Traitement parallélisé"):
future.result()
|