Expand description
Data Quality results as a report: what is likely wrong, what depends on intent, and
which columns have nothing to report. Each check kind is one row of
ObservationKind::spec (names, severity, rank, advice, grouping, wording). Built
once per result; the engine, cache and drill-in work on observations, which a
finding only groups and ranks.
Structs§
- Check
- One check the run makes: what it looks for, how far it reached, what it found. The list is the answer to “checked for what?” when nothing turned up.
- Check
Spec - One kind of check. Everything about a kind but its emitter and the predicate
beside it (
QualityObservation::evidence_predicate) is here. - Coverage
- How far the findings reach, beside them on every report: the checks that ran and what they read, the ones that did not and why, the rows behind the numbers, and what else bounds them. From what the run measured and saw; nothing here reads.
- Finding
- Findings
View - How the findings list is narrowed and ordered. Presentation only: it reads the report on screen, and nothing is measured again.
- Quality
Report - Report
Cache - The report and checks a
DataQualityResultsreads as, built once on first ask and kept with the results (their inputs never change after install). A copy starts empty, so a changed copy reads fresh.
Enums§
- Evidence
- What Enter opens of a finding, and how many rows that is.
- Evidence
Rows - The rows behind a finding.
- Finding
Order - The order findings are listed in within each severity.
- Grouping
- Which of a kind’s observations one finding says together.
- Outcome
- What one check found: nothing, something, nothing because there was nothing for it to look at, or nothing because this run could not look.
- Severity
- Variant
- A kind read more than one way (how much is missing; what parsing text holds), each with its own title and sometimes severity, rank and advice.
Constants§
- CHECKS_
SHOWN - How many checks the collapsed list shows before “more”.
- INTENT_
CHECK - The check the declared intent makes, by the name the Checks list gives it.
- NO_
FINDINGS - The clean-columns entry’s title.
- THIN_
SEGMENT_ ROWS - A segment with fewer sampled rows than this is thin: only a large change in it clears the sampling noise, and a clean one says little.
Functions§
- advice
- What a finding means for the data and what to do about it, a fragment a line.
- build_
report - check_
choices - The checks that made a finding in
report, each with how many, most important first. - checks
- Every check, most important first.
- column_
choices - The columns a report can be narrowed to, each with how many findings name it, in the results’ column order.
- columns_
label - Column names joined until
width, then “+N” for the rest. - coverage
- describe
- The finding’s numbers in a fragment, then the evidence that makes it concrete: the values, the spellings, the files.
- verdict
- Headline counts, one line: the answer before the evidence.