L'abstraction a toujours un coût qu'il est important de considérer. Abstraire exagérément et systématiquement c'est courir à la catastrophe.

L'abstraction a toujours un coût qu'il est important de considérer. Abstraire exagérément et systématiquement c'est courir à la catastrophe.

Using Google bq CLI, the following command allows to get the top Pypi keywords from the bigquery-public-data.pypi.distribution_metadata table:

bq query --use_legacy_sql=false 'SELECT keyword, COUNT(*) as keyword_count FROM `bigquery-public-data.pypi.distribution_metadata`, UNNEST(SPLIT(keywords, ", ")) as keyword GROUP BY keyword ORDER BY keyword_count DESC LIMIT 100'

Result for the top-15 keywords:

• python : 128555 appearances
• DuckDB Database SQL OLAP : 70739 appearances
• ai : 64997 appearances
• tensorflow tensor machine learning : 51144 appearances
• pulumi : 50076 appearances
• api : 47986 appearances
• probabilities probabilistic-graphical-models inference diagnosis : 46552 appearances
• rust : 45607 appearances
• cli : 39512 appearances
• OpenAPI : 38814 appearances
• sdk : 38060 appearances
• llm : 37487 appearances
• OpenAPI-Generator : 36734 appearances
• database : 35578 appearances
• automation : 34393 appearances

Note that this is a very basic query, that does take into account that some packages have a lot more versions published on Pypi than others.