Day-first vs month-first dates

finetype's datetime format reader treats two digits, a slash, two digits, a slash and four digits as one shape that is either dmy_slash or mdy_slash, and the same shape with dashes as either dmy_dash or mdy_dash.^[1] It counts a value towards that shape only when both two-digit fields are between 1 and 31 and the year is between 1000 and 9999.^[2][3] A value whose first field is over 12 and whose second is 12 or under is evidence that the day comes first, and a value with those fields the other way round is evidence that the month comes first.^[4] The reader names a type only when at least 95% of the column's sampled non-null values fit the shape and all of the evidence points one way.^[5][6] When no value has a field over 12, or some values point each way, the reader returns nothing; finetype's source says it never guesses the locale.^[7][8]

finetype applies the reader's choice only when at least 90% of the column's non-empty sampled values pass the chosen type's own validation rule.^[9][10] Earlier in the same pipeline, a value rule takes a column finetype's model has labelled dmy_slash or mdy_slash and changes the label to dmy_slash when some value's first field is over 12 and no value's second field is, or to mdy_slash the other way round.^[11][12][13] When neither field ever goes over 12, or both do, that rule leaves the model's label as it was, and when the reader returns nothing the reader's step leaves the column's type as it was too.^[14][15]

finetype's own tests pin 31/01/2020 with 05/06/2019 as dmy_slash, 01/31/2020 with 06/05/2019 as mdy_slash, and both 01/02/2020 with 03/04/2019 and 31/01/2020 with 01/31/2020 as read by neither.^[16]

A column in which no day is over 12 cannot be settled from its values, so its order has to come from whoever wrote the file, its documentation or its country of origin. (Meridian's view)

The evaluation report in finetype's repository at release 0.6.66 measured release 0.6.60 on 2026-09-23.^[17] The report says its tables are generated by a separate tool that has drifted from the SQL harness it claims to mirror, and that a figure taken from it for publication must be its SQL headline.^[18] Until finetype publishes a measurement for each type, this page gives no figure for how often it mistakes one order for the other. (Meridian's view)