From 0440409f4edce1702775fc3fc3ba1818f7d99eb7 Mon Sep 17 00:00:00 2001
From: Bastien Le Querrec <blq@laquadrature.net>
Date: Sat, 13 Apr 2024 17:05:24 +0200
Subject: [PATCH] RAAspotter: sauvegarde la date de modification des RAA
---
RAAspotter.py | 35 ++++++++++++++++++++++++++++++-----
1 file changed, 30 insertions(+), 5 deletions(-)
diff --git a/RAAspotter.py b/RAAspotter.py
index 98d3092..add05ee 100644
--- a/RAAspotter.py
+++ b/RAAspotter.py
@@ -45,6 +45,7 @@ class RAAspotter:
name = ""
sha256 = ""
pdf_creation_date = None
+ pdf_modification_date = None
def __init__(self, url, date, name):
if not url == "":
@@ -60,20 +61,39 @@ class RAAspotter:
self.sha256 = hashlib.sha256(self.url.encode('utf-8')).hexdigest()
return self.sha256
- def get_pdf_creation_date(self, data_dir):
+ def get_pdf_dates(self, data_dir):
raa_data_dir = f'{data_dir}/raa/'
+ p_pdf = None
+ pdf_parser = None
+ pdf_creation_date_raw = None
+ pdf_modification_date_raw = None
+
try:
p_pdf = open(f'{raa_data_dir}{self.get_sha256()}.pdf', 'rb')
pdf_parser = PDFParser(p_pdf)
+ except Exception as exc:
+ logger.warning(f'Impossible d\'ouvrir le PDF {self.get_sha256()}.pdf pour extraction de ses métadonnées : {exc}')
+
+ try:
pdf_creation_date_raw = PDFDocument(pdf_parser).info[0]['CreationDate'].decode('utf-8').replace('D:', '').replace('\'', '')
if pdf_creation_date_raw:
try:
self.pdf_creation_date = datetime.datetime.strptime(pdf_creation_date_raw, '%Y%m%d%H%M%S%z')
- except ValueError as exc:
+ except ValueError:
self.pdf_creation_date = datetime.datetime.strptime(pdf_creation_date_raw, '%Y%m%d%H%M%S')
except Exception as exc:
- logger.warning(f'Impossible d\'extraire la date du PDF {self.get_sha256()}.pdf')
+ pass
+
+ try:
+ pdf_modification_date_raw = PDFDocument(pdf_parser).info[0]['ModDate'].decode('utf-8').replace('D:', '').replace('\'', '')
+ if pdf_modification_date_raw:
+ try:
+ self.pdf_modification_date = datetime.datetime.strptime(pdf_modification_date_raw, '%Y%m%d%H%M%S%z')
+ except ValueError:
+ self.pdf_modification_date = datetime.datetime.strptime(pdf_modification_date_raw, '%Y%m%d%H%M%S')
+ except Exception as exc:
+ pass
def extract_content(self, data_dir):
raa_data_dir = f'{data_dir}/raa/'
@@ -97,22 +117,27 @@ class RAAspotter:
raa_data_dir = f'{data_dir}/raa/'
pdf_creation_date_json = None
+ pdf_modification_date_json = None
+
if self.pdf_creation_date:
pdf_creation_date_json = self.pdf_creation_date.strftime("%d/%m/%Y %H:%M:%S")
+ if self.pdf_modification_date:
+ pdf_modification_date_json = self.pdf_modification_date.strftime("%d/%m/%Y %H:%M:%S")
properties = {
'name': self.name,
'date': self.date_str,
'url': quote(self.url, safe='/:'),
'first_saw_on': datetime.datetime.today().strftime("%d/%m/%Y %H:%M:%S"),
- 'pdf_creation_date': pdf_creation_date_json
+ 'pdf_creation_date': pdf_creation_date_json,
+ 'pdf_modification_date': pdf_modification_date_json
}
f = open(f'{raa_data_dir}{self.get_sha256()}.json', 'w')
f.write(json.dumps(properties))
f.close()
def parse(self, data_dir, not_before, keywords):
- self.get_pdf_creation_date(data_dir)
+ self.get_pdf_dates(data_dir)
self.write_properties(data_dir)
self.extract_content(data_dir)
--
GitLab