Feature requests: more flexibly find date substring in a non-date string; and process additional incomplete date substrings about anytime

[eddelbuettel]

It's tricky. I think the 'leading stuff' maybe be best handled by an added 'filter' you could add:

R> anydate(gsub("(^[a-z ]*)", "", "asfdsfsdf 2020-01-01"))
[1] "2020-01-01"
R>

The second one is a no-no as our parsers aim to create proper date or datetime objects. And a pair on month and year is simply not a date. So there too you may have to build something on top of what we offer.

[chrisumphlett]

Thanks. The regex is a simple workaround for that one scenario. It doesn't help in two sub-cases I have:

1. Leading numbers in the string before the date
2. Date sub-strings with a format like "May 20, 2020".

As far as the need for proper objects, admittedly I'm far from an expert on how the parsers work. My suggestion is not that you only have a month and year, but that you have a simple set of rules to impute missing parameters (day =1, month/year = current month/year).

[eddelbuettel]

The package makes no claim to help with any possible date formats. Some remain too irregular. There is no way around this.

It does offer you ways to accomodate the need for special treatment, see help(addFormats).

I still think we should close this.

[chrisumphlett]

thanks

[eddelbuettel]

BTW your last example already works:

R> anytime::anydate("May 20, 2020")
[1] "2020-05-20"
R> 

[chrisumphlett]

That one works, with no leading text. This doesn't work:

anydate(gsub("(^[a-z ]*)", "", "asfdsfsdf May 20, 2020"))

[eddelbuettel]

Regular expressions are very powerful. It pays off to read up a little. Here [a-z] means letters; use [a-z0-9] and you may need to add punctuation etc -- this is no longer an anytime issue.

Best of luck, and I hope you continue to find anytime useful, even with challenging data.

[chrisumphlett]

addFormats is promising for me. Thank you for sharing that. Is the addition of a format permanent, just for the session, something else?

anytime::anydate("May 2020")
[1] NA
anytime::addFormats("%b %Y")
anytime::anydate("May 2020")
[1] "2020-05-01"

[eddelbuettel]

Just for the session. But nothing is stopping you from adding library(anytime); addFormats(c("....". "...")) to your R startup. Or maybe to the scripts or functions you use for this project.

[chrisumphlett]

Same logic also for date times. anytime doesn't like HH:MM with no :SS. getFormats always has %H:%M:%S. And then anytime also doesn't like am/pm either...? I don't see that in any of the formats. I'm opening a separate thread about time zones.

[chrisumphlett]

Just for the session. But nothing is stopping you from adding library(anytime); addFormats(c("....". "...")) to your R startup. Or maybe to the scripts or functions you use for this project.

Yep understood, just making sure I knew if it was necessary. I'd probably to it on a project/process-level. I use anytime in almost every project and I've never needed to alter the formats.