[C++] Is CSV reader's TimestampParser usable elsewhere?

[asfimport]

The TimestampParser seems to be able to cycle through several formats. This sort of functionality would be very useful for some of the lubridate bindings that need to behave in a similar way.

library(arrow)
library(fs)
library(readr)
library(tibble)

tf <- fs::file_temp(ext = "csv")
fs::file_create(tf)

sample_times <- tibble(a = c("09/13/2013", "25/12/1998", "09-13-13", "23_Feb_2022", "09/13/2018"))
write_csv(sample_times, tf)


read_csv_arrow(tf, 
               as_data_frame = TRUE,
               timestamp_parsers = c("%m/%d/%Y", "%d/%m/%Y", "%m-%d-%y", "%d_%b_%Y"))
#> # A tibble: 5 × 1
#>   a                  
#>   <dttm>             
#> 1 2013-09-13 01:00:00
#> 2 1998-12-25 00:00:00
#> 3 2013-09-13 01:00:00
#> 4 2022-02-23 00:00:00
#> 5 2018-09-13 01:00:00

For example, in lubridate, the ymd() cycles through all possible formats that have year-month-date components in the right order (e.g. "%Y-%m-%d", "%y-%m-%d", "%Y-%b-%d", "%y-%b-%d", "%Y-%B-%d", "%y-%b-%d", etc).

I guess my question is: Can we factor this CSV reader feature to be usable elsewhere? This was the bit that caught my attention: "using the virtual parser interface in arrow/util/value_parsing.h", and told me that using it elsewhere might be a possibility.

Reporter: Dragoș Moldovan-Grünfeld / @dragosmg

Related issues:

• [C++] Strptime issues umbrella (is a child of)

_{Note: This issue was originally created as ARROW-15912. Please see the migration documentation for further details.}

[asfimport]

Antoine Pitrou / @pitrou:
Well, the timestamp parser functionality is available in arrow/util/values_parsing.h. You'll have to reimplement the logic to loop through parsers yourself, but that should be close to trivial.

Does this issue need to be kept open?