Chart pages of a report
--details writes, beside the report, the pages of each column and the chart
pages: interactive charts of what the second pass counted exactly. The first one
draws the combinations of columns the records lack a value in.
tabalyst report customers.csv --detailscustomers.htmlcustomers/charts-missing.htmlcustomers/col-01-id.html...--details runs the second pass as --deep does: it asks first
when it is costly, --yes skips the question, and outside a terminal a costly pass
stops with exit code 3 without --yes. With {"deep": {"mode": "off"}} in the
configuration, --details is an error.
Missing values: charts-missing.html
Section titled “Missing values: charts-missing.html”One section per dataset of the report, with the anchor of the dataset
(charts-missing.html#rows for a CSV file). Each section shows:
- Three figures: the records with every value, those that lack at least one, and the combinations seen.
- The blocks: the largest groups of columns that go missing together, the same sentences as the card of the report, for example:
date_reclamation, montant_reclamation and reclamation_status go missing together: 85,760 records lack all three, at least 97.1% of the records that lack any one of them.- An UpSet plot. The bars at the top count the records of each combination; the dots below name its columns; the bars on the left count the records that lack each column. Each record is in one combination only, so the bars at the top add up to the records that lack a value. The plot draws the first 20 combinations; sort them by records or by number of columns, and show every combination, those of two columns or more, or those that hold one column. On a narrow screen it becomes one bar per combination.
- The tables: every combination the second pass lists, with its columns, records and share, then the columns that lack values, each linked to its column page. The tables hold every figure of the plot.
A combination is the exact set of columns a record lacks a value in: a record that
lacks a phone and a reference and has every other value is in the combination
phone, ref, not in phone nor in ref. Missing means an empty cell, a blank one
or a null marker of the configuration, as the scan counts it. The counts are exact.
When a file has more combinations than deep.missing.max_patterns, the records of
the later ones are counted together and the page says so. See Combinations of
missing values.
The report links to the page from the Columns missing together card of the dataset and from its navigation; a column page links to it when the column lacks values. No page is written when no dataset has two columns that lack values; the command says why:
Warning [customers.csv]: no chart of the missing values for rows: fewer than two columns lack a valueA library from a CDN
Section titled “A library from a CDN”The chart pages draw with Apache ECharts, loaded from jsDelivr. The address names one version and the page checks the file against its hash, so it cannot be swapped; no data of the report is sent, and the page asks the browser not to send its address. The report itself and the column pages load nothing from outside and read without a network.
Without a network, or behind a proxy that blocks the CDN, each chart shows a message in its place and the tables stay: they hold the same figures.