IP code (IEC 60529) ingress-protection ratings as three CSV tables

<p dir="ltr">The IP code is read as a scale and it is not one. IPX7 does not imply IPX5: a device certified for thirty minutes under a metre of still water can fail a garden hose, because the two digits describe unrelated tests and only the second one was run. The letter X is not a zero either — it records that the digit was never tested, so IPX7 says nothing at all about dust.</p><p dir="ltr">This dataset is the <a href="webstore.iec.ch/publication/2452" target="_blank">IEC 60529</a> ingress-protection code split into three CSV tables so the distinction can be checked programmatically rather than read off a marketing page.</p><ul><li><b>ip_code_first_digit.csv</b> — 7 rows, solids protection 0–6, with the diameter of the standard test probe for each.</li><li><b>ip_code_second_digit.csv</b> — 10 rows, water protection 0–8 plus the 9K test from <a href="iso.org/standard/45555.html" target="_blank">ISO 20653</a>, with nozzle, flow and duration per digit.</li><li><b>ip_code_common_marks.csv</b> — 11 rows, the marks that actually appear on products (IPX4, IP67, IP69K…), each with a <code>what_it_does_not_tell_you</code> column.

That column is the point of the dataset.</li></ul><p dir="ltr">Two design choices are worth stating. Digit 8 carries no fixed depth or duration because the standard leaves both to the manufacturer, which is why two IP68 devices are not comparable to each other. And <code>survives_jets</code> is recorded as <i>unknown</i>, never as <i>no</i>, for marks ending in 7 or 8 — the jet test was not run, and absence of evidence is stored as absence of evidence.</p><p dir="ltr">Transcribed from the publicly documented test conditions; <a href="iec.ch/ip-ratings" target="_blank">IEC's own summary of the rating system</a> and the <a href="en.wikipedia.org/wiki/IP_code" target="_blank">Wikipedia IP code article</a> are the easiest cross-checks against the paywalled standard.