Comment by onceonceonce

2 hours ago

been hand-rolling date -d plus a wrapper script for newsletter scheduling for years. Stuff like "next second Tuesday after a holiday" and "convert these timestamps to a reader's local time before send". biff time seq monthly -w 2-tue would have replaced about 40 lines of bash for me.

2 comments

onceonceonce

burntsushi 2 hours ago

Yes!!! To spell it out (with the new biff -> bttf name):

    $ bttf time seq monthly -w 2-tue -u 1y | bttf time fmt -f '%c'
    Tue, Jun 9, 2026, 11:15:11 AM EDT
    Tue, Jul 14, 2026, 11:15:11 AM EDT
    Tue, Aug 11, 2026, 11:15:11 AM EDT
    Tue, Sep 8, 2026, 11:15:11 AM EDT
    Tue, Oct 13, 2026, 11:15:11 AM EDT
    Tue, Nov 10, 2026, 11:15:11 AM EST
    Tue, Dec 8, 2026, 11:15:11 AM EST
    Tue, Jan 12, 2027, 11:15:11 AM EST
    Tue, Feb 9, 2027, 11:15:11 AM EST
    Tue, Mar 9, 2027, 11:15:11 AM EST
    Tue, Apr 13, 2027, 11:15:11 AM EDT
    Tue, May 11, 2027, 11:15:11 AM EDT

More examples here: https://github.com/BurntSushi/biff/blob/master/GUIDE.md#date...

Implementing the RFC 5545 recurrence rules was quite a lot of fun: https://github.com/BurntSushi/biff/blob/4c75d5cf6e09310e74ca...

I'm quite proud of it, because if you look at the implementation, it's almost entirely about dealing with the specification rules. All of the datetime bullshit (including handling time zones) is all deferred to Jiff.

Plus, the tests are nearly 4,000 lines. While the implementation is 2,000 lines.

jayknight 1 hour ago

I implemented those recursion rules (probably very poorly) years ago. I still often think about the ways I did and probably should have handled those. It was an interesting problem.