README.md

arxiv-browse

Running Browse with the Flask development server

You can run the browse app directly.

make venv

(the make rarely works)

or

python --version
# 3.11.x
python -m venv ./venv
source ./venv/bin/activate
pip install poetry==1.3.2
poetry install
python main.py

Note -- make sure you have python dev installed befoore doing the above steps, or the poetry install will fail trying to build the mySQL library dependency. E.g.: sudo apt-get install python3.11-dev

Then go to http://127.0.0.1:8080/abs/0906.5132

This will monitor for any changes to the Python code and restart the server. Unfortunately static files and templates are not monitored, so you'll have to manually restart to see those changes take effect.

By default, the application will use the directory trees in tests/data/abs_files and tests/data/cache and when looking for the document metadata and PDF files. These paths can be overridden via environment variables (see browse/config.py).

This method will only give you access to minimal data in the built-in test dataset. For full (read-only) access to the production database, the following steps are necessary:

Running with access to the production database

Prerequisites: you are logged into gcloud and have default application credentials. This can be achieved by calling gcloud auth login --update-adc --force and logging into your @arxiv.org account.

Next, all the following steps have to be done.

Also, in the following we assume that you two port variable set in the environment:

MAIN_DB_PORT=3301
LATEXML_DB_PORT=3302

(the numbers are not important, but must be different)

Create a .env file

First, you'd need to create the '.env' file somewhere. Using tests/.env is suggested.

DOCUMENT_ABSTRACT_SERVICE=browse.services.documents.db_docs
ABS_PATH_ROOT=gs://arxiv-production-data
DOCUMENT_CACHE_PATH=gs://arxiv-production-data/ps_cache
DOCUMENT_LISTING_PATH=gs://arxiv-production-data/ftp
DISSEMINATION_STORAGE_PREFIX=gs://arxiv-production-data
LATEXML_ENABLED=True
LATEXML_BUCKET="gs://latexml_document_conversions"
LATEXML_BASE_URL=//127.0.0.1:8080
FLASKS3_ACTIVE=1
CLASSIC_DB_URI=" SEE BELOW "
LATEXML_DB_URI=" SEE BELOW "

The value of CLASSIC_DB_URI one can obtain by

MAIN_SECRET=$(gcloud secrets versions access latest --project=arxiv-production  --secret=readonly-arxiv-db-uri)
echo $MAIN_SECRET | sed -e "s#/arXiv.*#127.0.0.1:${MAIN_DB_PORT}/arXiv#"

The value of LATEXML_DB_URI one can obtain by

LATEXML_SECRET=$(gcloud secrets versions access latest --project=arxiv-production  --secret=latexml_db_uri_psycopg2)
echo $LATEXML_SECRET | sed -e "s#/latexmldb.*#127.0.0.1:${LATEXML_DB_PORT}/latexmldb#"

If you have a PyCharm, script: main.py Enable env files Add tests/.env

Running cloud-sql-proxy

You can obtain the database name MAIN_DB_NAME from the same $SECRET as in the previous step:

MAIN_DB_NAME=$(echo $MAIN_SECRET | sed -e "s#^.*unix_socket=/cloudsql/##")

Similar, for the LATEXML_DB_NAME one needs to do

LATEXML_DB_NAME=$(echo $LATEXML_SECRET | sed -e "s#^.*host=/cloudsql/##")

NOTE: cloud_sql_proxy and cloud-sql-proxy (new) have different options.

cloud-sql-proxy --address 0.0.0.0 --port ${MAIN_DB_PORT} ${MAIN_DB_NAME}
cloud-sql-proxy --address 0.0.0.0 --port ${LATEXML_DB_PORT} ${LATEXML_DB_NAME}

For the old version

cloud_sql_proxy --instances=${MAIN_DB_NAME}=tcp:${MAIN_DB_PORT}
cloud_sql_proxy --instances=${LATEXML_DB_NAME}=tcp:${LATEXML_DB_PORT}

If the proxy is working, you can use mysql client to connect to the db.

mysql -u browse -p --host 127.0.0.1 --port ${MAIN_DB_PORT} arXiv
Enter password: 
...
Type 'help;' or '\h' for help. Type '\c' to clear the current input statement.

mysql> show tables;
+------------------------------------------+
| Tables_in_arXiv                          |
+------------------------------------------+
| Subscription_UniversalInstitution        |

Run the main server

Run

python main.py`

should start up and run

Test suite

Run the main test suite with the following command:

pytest tests

Running Browse in Docker

You can also run the browse app in Docker. The following commands will build and run browse using defaults for the configuration parameters and will use the test data from tests/data. Install Docker if you haven't already, then run the following:

script/start

This command will build the docker image and run it. If all goes well, http://localhost:8000/ will render the home page.

Configuration Parameters

See browse/config.py for configuration parameters and defaults). Any of these can be overridden with environment variables.

Serving static files on S3

We use Flask-S3 to serve static files via S3.

After looking up the AWS keys and region and bucket:

cd arxiv-browse
git pull
AWS_ACCESS_KEY_ID=x AWS_SECRET_ACCESS_KEY=x \
 AWS_REGION=us-east-1 FLASKS3_BUCKET_NAME=arxiv-web-static1 \
 pipenv run python upload_static_assets.py

In AWS -> CloudFront, select the static.arxiv.org distribution, -> Invalidations -> Create invalidation, and enter a list of url file paths, eg: /static/browse/0.3.4/css/arXiv.css.

It may be help to use a web browser's inspect->network to find the active release version.

Tests and linting for PRs

There is a github action that runs on PRs that merge to develop. PRs for which these tests fail will be blocked. It is the equivalent of running:

# if there are Python syntax errors or undefined names
flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics

pytest tests

Setting up pytest in PyCharm