Is this the worst date format ever?

[andrewducker]

I think I just found the worst date format ever.

CVS, in its "Entries" file stores the timestamp that the file was last changed, so that you can check if it has been updated.

The timestamp is formatted like this: "Mon Dec  7 14:59:26 2010"

Now, to a human that's a perfectly readable format. But if you want a computer to understand it you've got to separate it into its component parts which are _completely_ out of order.

ISO 8601 is the international standard. And following that, the format would be "2010-12-20T14:59:26Z". Which is slightly harder to read as a human (because it's all smooshed together), but so much easier to parse with a computer. And has the bonus that if you sort a bunch of dates using normal character order they'll come out correctly.

CVS gets a bit of an out for being created in 1986, 2 years before ISO 8601. But as ISO 3307 was available from 1975, there were standards they totally _could_ have used.

(I can't find details on what the ISO 3307 date-time format looked like, but ISO 2014 said to use YYYY-MM-DD, so I can only assume that ISO 3307 did something reasonably sensible.)

Next task for today: Write a parser for that date format.

(And look forward to a glorious future when I won't have to touch CVS any more.)

Comments

[jack]

Ugh. There's got to be a "worst date" joke in here somewhere. And for that matter, there's surely something worse, where the dates are utterly ambiguous, or something, but probably not widely used.

[redbird]

My only thought here is something you're probably already thought of: you can throw away the "Mon" part as redundant before parsing the rest of it.

Edit after looking at Jack's comment: I suppose if a system was storing dates as "Mon 12/7 14:59:26 2010" (instead of "Dec 7") the "Mon" might be useful for determining whether that meant December 7th or July 12th. But that seems like an unlikely choice even for human readability.

[momentsmusicaux]

You still have to use CVS?????

[andrewducker]

Currently working to get it off it, but it's a massive project. Huge amounts of things built on top of it.

[momentsmusicaux]

Git became mainstream about 10 years ago. SVN probably 15 years ago.

That's a LONG time to put off getting off a shitty system.

[andrewducker]

Putting together a business case to spend a lot of money on something sure the business benefits aren't easy to quantify is...tricky.

There are always things to spend money on, getting to the top of the pile is the hard part.

[ckd]

ctime() format, ugh. You can't really blame CVS for anything except laziness, though, since it's a library function.

[ckd]

Also, to parse it you can probably use the ancient and crappy strptime() which dates to the same era as ctime()/asctime() (surprised? didn't think so).

[j_v_lynch]

We've been saying "Next release we're switching from CVS" for probably 4 years now.

[momentsmusicaux]

Also:

> But if you want a computer to understand it you've got to separate it into its component parts which are _completely_ out of order.

Nope.

$ php -r "print strtotime('Mon Dec 7 14:59:26 2010');"
1292252366

Done!

[birguslatro]

"Next task for today: Write a parser for that date format." ...

>> date: load "Mon Dec 7 14:59:26 2010"
== [Mon Dec 7 14:59:26 2010
]
>> foreach value date [print rejoin [type? value " " value]]
word Mon
word Dec
integer 7
time 14:59:26
integer 2010
>> date/4/hour
== 14
>> date/4/minute
== 59
>> date/4/second
== 26.0

:)

[andrewducker]

DateTime.ParseExact(timestamp, "ddd MMM d HH:mm:ss yyyy", null, DateTimeStyles.AllowInnerWhite);

(I'm fairly sure there's a Date Parse method that would be slightly simpler, but I had other things to do)

[birguslatro]

My example's in Rebol and your datestamp just happened to be correct Rebol code. So...

load datestamp

was all that was needed. It shows the advantage of having a rich set of datatypes coupled with whitespace separators. So, human readable and (near enough to) computer readable.

[andrewducker]

I suspect REBOL was written in C and so supports the ancient C standard time format.

[birguslatro]

No. In Rebol, time is a datatype, same as integers and strings and such are datatypes. Which allows you to do this kind of thing with them...

>> 10:3:17 + 3:2:1
== 13:05:18

Date is also a datatype, and your one would look like this: 7-Dec-2010/14:59:26

So...

>> date: 7-Dec-2010/14:59:26
== 7-Dec-2010/14:59:26
>> date/year
== 2010
>> date/month
== 12
>> date/weekday
== 2

Which suggests your Mon is wrong. Rebol's weekdays start at 1 for Monday.