Handle dates like "2001" and "2001-10" better

[zostay]

If you enter a partial date like the above, it will sort of work, but it does so by falling back to Date::Manip. When that happens, it assumes you mean only a single day.

Thus, "2001" becomes:

# Got:
[ "2001-01-01 00:00:00", "2001-01-01 23:59:59" ]

# Expected:
[ "2001-01-01 00:00:00", "2001-12-31 23:59:59" ]

and "2001-10" becomes:

# Got:
[ "2001-10-01 00:00:00", "2001-10-01 23:59:59" ]

# Expected:
[ "2001-10-01 00:00:00", "2001-10-31 23:59:59" ]

These dates ought to be handled more naturally.

[zostay]

Looks like Date::Manip has some tools that are helpful in solving this problem correctly. It appears to be only available under DM6, but that's good enough for me.

We can detect which fields were given and which were implied during the parsing of the date. We can then use that to determine what the range should be.

See:

• complete method of Date::Manip::Date
• Complete vs. truncated dates and times

[zostay]

Or maybe it's completely useless. It seems to return that all fields are complete even when I just give it a year. I'm investigating whether I'm just doing it wrong or if there's a bug in the version Date::Manip I'm using.

[mjaquilina]

My general inclination is to avoid relying on Date::Manip as much as possible, except as the final fallback method. Not sure if that's for legitimate reasons or just NIH.

[zostay]

Yes, but if we fall back to it, it would be nice to do the right thing. Based on my experience here, the "complete" method does not act like the documentation says. I'm probably going to abandon the effort.

Speaking of NIH, ever heard of DateTime::Format::Natural?