Everything useful before you download or buy.
A check for each dimension of data quality, each with its real failing count.
Two A4 pages of SQL checks, written on the agricultural tables of Can I Trust This Data?: exports, prices and farm technology, a teaching copy with faults put in on purpose. Every query was run, and its real result is printed under it.
It starts with one profiling query that works on any table (NULLs, blanks and padding, distinct values, and spellings that differ only in case), then gives the checks for each dimension: missing values and missing rows counted against the rows that should exist, uniqueness on a key with the three counts people mix up, ranges, formats and dates after the as-of date, consistency with master data and between tables, and the age of the data on a fixed as-of date.
It ends with one way to report every check (rule, rows checked, rows failing, pass rate) and the NULL trap that lets a missing value pass a rule without a word.
PDF to print, and an HTML version to read on screen. The SQL is DuckDB, as in the LucidPlexus workbench, with short notes on what to write in PostgreSQL, SQL Server or MySQL where a query depends on DuckDB.
Check the resource before buying — a limited-page preview.
This is a limited-page preview. If it does not load, open the PDF directly.
One review per person. Submitting again updates your existing review.
Be the first to review this resource.