scaper-place

scraper-place est un scraper pour le site internet gouvernemental https://www.marches-publics.gouv.fr/ (PLACE pour PLateforme d'AChat de l'Etat). scraper-place n'a rien à voir avec Jean-Vincent Placé.

PLACE permet d'accéder aux Dossiers de Consultation des Entreprises (DCE) pour les consultations en cours. Malheureusement, PLACE ne permet pas un accès pratique aux ressources publiées pour tous les usages. D'une part, les dossiers des publications passées sont retirées. D'autre part, le fonctionnement du site rend malaisée le requetage automatisé des DCE.

scaper-place va récupérer tous les DCE (ainsi que les métadonnées associées) pour les répliquer localement et les indexer.

Le droit d'accès applicable aux DCE est résumé dans une fiche de la CADA.

How PLACE works, how to parse it

PLACE uses the PHP framework PRADO. It stores the current navigation state in the PRADO_PAGESTATE, an encoded variable of about 100kB. This variable is required (along with PRADO_POSTBACK_TARGET and optional parameters) to perform a request. There is no way to get a list of all the available DCE. The search engine is clearly not an option. The most convenient way I found is to request the paginated list of all the current consultations, set the pagination number to 20 (the maximum) and request all the pages in order. Naturally, given the id of a consultation, three successive requests are needed to access the document.

Curiously, a small fraction of the DCE appear in several pages, and this is not related to the addition of documents during the course of the parsing. I guess such a feature would be very difficult to implement purposefully.

Features

• scrap PLACE every night
• text extraction using Apache Tika
• indexation with ElasticSearch
• backup of both documents and metadata on AWS S3

Install scraper-place

Prerequisites

• Install mongodb 6 (other versions may work).
• If you plan to replicate the files on AWS Glacier, create a vault and create a IAM user with upload permission.
• If you plan to index the data with ElasticSearch, install it.
• Create a python virtual env with python>=3.9 (I suggest using pew).

Installation

• Clone this repository.
• In the repository directory: pip install --editable .
• Copy config.ini.example to config.ini and set your configuration.
• Create the directories you configured in config.ini and make sure they are writable by the process that will run scraper-place.
• Import metadata to mongo (see scripts/import-to-mongo.ipynb)
• Configure ElasticSearch (see elasticsearch.yml)
• Set up the ElasticSearch index (see scripts/create_index.ipynb)
• Setup services (see betterplace.service, tika.service)
• Configure nginx (see betterplace.info)
• Setup crons to trigger scripts/nightly_scraping.sh and scripts/backup_metadata.py (see crontab for an example)

Misc

To use debug logging on elasticsearch:

curl -XPUT 'localhost:9200/_cluster/settings' --data '{"transient":{"logger._root":"DEBUG"}}' -H'Content-Type: application/json'