Add date_order locale option and flexible date separator parsing by hidekoji · Pull Request #623 · tidyverse/vroom

hidekoji · 2026-05-18T07:01:24Z

Summary

Adds a date_order argument to locale() and makes date / date-time auto-detection more forgiving, so columns of year-last dates (e.g. 10/02/2024) can be read as dates instead of being guessed as character.

locale(date_order =) — new optional argument accepting an explicit component order: "ymd", "mdy", "dmy", etc., optionally with a time suffix ("mdy_hms", "dmy_hm", "ymd_h"). NULL (default) keeps the current automatic behaviour. Validated in R with a clear cli_abort() message.
DateTimeParser::parseDateOrder() — parses a value against an explicit order, including an optional T/space-separated time part.
DateTimeParser::parseYearLastHeuristic() — recognises unambiguous D/M/YYYY vs M/D/YYYY (part > 12 disambiguates; defaults to MDY when ambiguous, the US convention). Used as an auto-detection fallback in isDate() / parse_date().
Flexible separators — parseISO8601() / parseDate() now accept any non-alphanumeric separator between date components (2024.10.02, 2024/10/02, …), similar to lubridate's ymd().
parse_date() / parse_dttm() now receive the LocaleInfo* so they can honour date_order.

When date_order is set, guess_type() routes date-only orders to isDate() and time-suffixed orders to isDateTime(), and will not cross-match the other kind.

Test plan

tests/testthat/test-datetime.R gains end-to-end vroom() coverage:

explicit date_order for MDY / DMY dates and mdy_hms / dmy_hms date-times
auto-detection of unambiguous DMY year-last dates without date_order
ambiguous year-last dates default to MDY
NA handling and dot/slash/dash separator variants
existing YMD / ISO-8601 behaviour unchanged (backward compatibility)

`locale()` gains a `date_order` argument so dates and date-times can be parsed with an explicit component order ("mdy", "dmy", "ymd_hms", etc.). This makes year-last formats such as 10/02/2024 readable, which the automatic type guesser would otherwise treat as character. Date and date-time auto-detection now also accepts any non-alphanumeric separator between components and falls back to a year-last heuristic that disambiguates D/M/YYYY vs M/D/YYYY (defaulting to MDY when ambiguous). Adds end-to-end vroom() tests covering explicit date_order, auto MDY/DMY detection, separator variants, and YMD backward compatibility.

utils::unzip(list=TRUE) on Windows R 4.2.x garbles non-ASCII entry names (e.g. ä → <84>), causing unz() to fail to locate the entry. When the archive package is installed, route zip reading through archive::archive_read() which handles UTF-8 entry names correctly. Preserves existing behaviour when archive is not installed (falls back to the utils::unzip / unz() path). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

…time heuristic The year-last date heuristic only accepted 4-digit years (it read the year with consumeInteger(4, exact=true) and rejected year < 1000), so values like 5/29/26 were guessed as character. Datetime auto-detection only handled ISO8601. - Add consumeYearFlexible(): consume a 2- or 4-digit year, applying the same pivot as the %y format specifier (00-68 -> 2000s, 69-99 -> 1900s) and rejecting implausible 3-digit values. Route parseYearLastHeuristic() and the parseDateOrder() year component through it. - Add parseYearLastHeuristicDateTime() (year-last date + T/space + HH[:MM[:SS]] with optional tz) and wire it into guess_type isDateTime() and vroom_dttm.cc materialization as an ISO8601 fallback, so MDY/DMY datetimes are recognized. - Extract disambiguateDayMonth() shared by both heuristics. - Tests for 2-digit MDY/DMY/ambiguous dates, the %y pivot, invalid/3-digit rejection, 2- and 4-digit MDY datetimes, and explicit date_order with 2-digit years.

github-actions · 2026-06-01T22:37:16Z

+  csv <- "id,date\n1,5/29/26\n2,5/31/26\n3,12/25/26"
+  result <- vroom::vroom(I(csv), show_col_types = FALSE)
+  expect_s3_class(result$date, "Date")
+  expect_equal(result$date, as.Date(c("2026-05-29", "2026-05-31", "2026-12-25")))


[air] _{reported by reviewdog 🐶}

Suggested change

expect_equal(result$date, as.Date(c("2026-05-29", "2026-05-31", "2026-12-25")))

expect_equal(

result$date,

as.Date(c("2026-05-29", "2026-05-31", "2026-12-25"))

)

github-actions · 2026-06-01T22:37:16Z

+
+test_that("vroom() does not treat invalid or 3-digit-year values as year-last dates", {
+  for (v in c("13/25/26", "100/200/300")) {
+    result <- vroom::vroom(I(paste0("x\n", v, "\n")), delim = ",", show_col_types = FALSE)


[air] _{reported by reviewdog 🐶}

Suggested change

result <- vroom::vroom(I(paste0("x\n", v, "\n")), delim = ",", show_col_types = FALSE)

result <- vroom::vroom(

I(paste0("x\n", v, "\n")),

delim = ",",

show_col_types = FALSE

)

github-actions · 2026-06-01T22:37:16Z

+})
+
+test_that("vroom() reads 2-digit-year dates with explicit date_order", {
+  res_mdy <- vroom::vroom(I("id,date\n1,5/29/26\n2,3/15/26"), locale = locale(date_order = "mdy"), show_col_types = FALSE)


[air] _{reported by reviewdog 🐶}

Suggested change

res_mdy <- vroom::vroom(I("id,date\n1,5/29/26\n2,3/15/26"), locale = locale(date_order = "mdy"), show_col_types = FALSE)

res_mdy <- vroom::vroom(

I("id,date\n1,5/29/26\n2,3/15/26"),

locale = locale(date_order = "mdy"),

show_col_types = FALSE

)

github-actions · 2026-06-01T22:37:16Z

+  expect_s3_class(res_mdy$date, "Date")
+  expect_equal(res_mdy$date, as.Date(c("2026-05-29", "2026-03-15")))
+
+  res_dmy <- vroom::vroom(I("id,date\n1,29/5/26\n2,15/3/26"), locale = locale(date_order = "dmy"), show_col_types = FALSE)


[air] _{reported by reviewdog 🐶}

Suggested change

res_dmy <- vroom::vroom(I("id,date\n1,29/5/26\n2,15/3/26"), locale = locale(date_order = "dmy"), show_col_types = FALSE)

res_dmy <- vroom::vroom(

I("id,date\n1,29/5/26\n2,15/3/26"),

locale = locale(date_order = "dmy"),

show_col_types = FALSE

)

hidekoji · 2026-06-01T22:39:24Z

Added a commit extending the year-last date heuristic to 2-digit years (e.g. 5/29/26), plus a year-last datetime heuristic:

consumeYearFlexible() consumes a 2- or 4-digit year and applies the standard %y pivot (00–68 → 2000s, 69–99 → 1900s), rejecting implausible 3-digit values; the year-last date heuristic and the date_order year component route through it.
parseYearLastHeuristicDateTime() recognizes M/D/Y / D/M/Y date + HH[:MM[:SS]] (optional tz); wired into the guesser and the materialization path so MDY/DMY datetimes are detected.
New tests: 2-digit MDY/DMY/ambiguous dates, the %y pivot, invalid/3-digit rejection, 2- and 4-digit MDY datetimes, and explicit date_order with 2-digit years. Existing date/datetime tests unchanged (incl. "%Y requires 4 digits" — explicit %Y/%y format paths are untouched). DESCRIPTION not bumped.

hidekoji force-pushed the feature/date-order-multi-format branch from 235789b to 4f2aa0a Compare May 18, 2026 07:11

hidekoji mentioned this pull request May 19, 2026

Add date_order locale option and flexible date separator parsing tidyverse/readr#1624

Open

5 tasks

hidekoji force-pushed the feature/date-order-multi-format branch from 4f2aa0a to 7ffee32 Compare May 19, 2026 19:23

hidekoji and others added 3 commits May 25, 2026 11:23

style: wrap long needs_archive line per lintr

c0b8ed4

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

github-actions Bot reviewed Jun 1, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add date_order locale option and flexible date separator parsing#623

Add date_order locale option and flexible date separator parsing#623
hidekoji wants to merge 4 commits into
tidyverse:mainfrom
hidekoji:feature/date-order-multi-format

hidekoji commented May 18, 2026

Uh oh!

github-actions Bot Jun 1, 2026

Uh oh!

github-actions Bot Jun 1, 2026

Uh oh!

github-actions Bot Jun 1, 2026

Uh oh!

github-actions Bot Jun 1, 2026

Uh oh!

hidekoji commented Jun 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

hidekoji commented May 18, 2026

Summary

Test plan

Uh oh!

github-actions Bot Jun 1, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot Jun 1, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot Jun 1, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions Bot Jun 1, 2026

Choose a reason for hiding this comment

Uh oh!

hidekoji commented Jun 1, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant