Someone sends you a CSV. Then a folder of CSVs. Then a CSV that's actually tab-separated but named .csv, with a stray header row and a column that's a number on most rows and the string N/A on the rest.
For years my answer to "can you pull a quick number out of this?" was a throwaway Python script. Read it in, fight pandas about dtypes, groupby, print, delete the script, forget everything, repeat next week. It worked. It was also slow and I never kept any of it.
These days I just point SQL at the file. I want to show you the exact workflow because it's embarrassingly simple and it's saved me a lot of evenings.
The one binary I actually use
The tool is clickhouse-local. It's a single binary, the ClickHouse engine minus the server. You download it and run SQL against files on your disk.
