parse_time(%M) fails to parse single digit minutes

[twest820]

This was reported on StackOverflow in August 2021 or so. I just hit it with stringr 1.5.0 on

readr::parse_time("9:0:0.057", format = "%H:%M:%OS")

From what I can tell, the essential issue is DateTimeParser.h hard codes %M as

      case 'M': // minute
        if (!consumeInteger(2, &min_))
          return false;
        break;

rather than testing the width of the minutes field (as is done for unrestricted hours) or using more flexible parsing (as is done for seconds). This is a departure from, say, %M in lubridate::parse_date_time(), which accepts both the 0–59 and 00–59 formats

Since I don't have access to the source code generating the .csv files being parsed here changing the data format isn't an option. One option for a workaround is to read the times into character column and then convert with lubridate::hms(), ms(), hm(), or similar.

[hadley]

x <- "1:2:3.057"

readr::parse_time(x, format = "%H:%M:%OS")
#> Warning: 1 parsing failure.
#> row col            expected    actual
#>   1  -- time like %H:%M:%OS 1:2:3.057
#> NA
strptime(x, format = "%H:%M:%OS")
#> [1] "2023-07-31 01:02:03 CDT"

^{Created on 2023-07-31 with reprex v2.0.2}