5
$\begingroup$

Bug introduced in 12.1 and persisting through 12.2.0

Update

WRI has confirmed that this is a bug. CASE:4521676


The performance of Min and Max on a list of DateInterval is O(2^n) where n is the length of the list.

$Version
(* 12.1.0 for Mac OS X x86 (64-bit) (March 14, 2020) *)

Generate some random date intervals

SeedRandom[11]; intervals = 30;

dateIntervals = 
 Table[DateInterval[{DateObject[{2019, 6, 15, 16, k = RandomInteger[{0, 59}]}], 
    DateObject[{2019, 6, 15, 16, RandomInteger[{k, 59}]}]}], intervals];

Timing Min for lengths 1 to 24. This takes ~5 min on my machine.

timings = Table[{n, dateIntervals // Take[#, n] & // Min // AbsoluteTiming // First}, {n, 1, 24}];

Fit 2^n

nlm = NonlinearModelFit[timings, b 2^n, {b}, n]

Show[ListLogPlot[timings], LogPlot[nlm[x], {x, 1, 24}]]

enter image description here

An easy workaround

dateIntervals // Map[Min] // Min

Strange that this simple operation would have performance issues.

Is this reproducible on other platforms? Any other workaround ideas?

I have reported this to WRI.

$\endgroup$
1
  • 4
    $\begingroup$ You can increase the performance of the workaround by replacing all the Min with MinDate. $\endgroup$ – Edmund Apr 2 '20 at 21:26
0
$\begingroup$
$Version

(* "12.0.0 for Mac OS X x86 (64-bit) (April 7, 2019)" *)

Result plot: enter image description here

Make use of the workaround with AbsoluteTime and the built-in Interval .

Hope that helps.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.