[1.0.1] Region exists(..., 0) failing does not return instantly with scan rates > 1 --- fixed in 1.1.0

[kurokrosk]

Version: 1.0.1
64-bit

Procedure: I look for a Pattern myPattern that is NOT in the Region myRegion. The size of the pattern (+similarity) and the region are irrelevant.
Expected behavior: The exists/find should return immediately.
Observed behavior: That is not always the case, there is some delay (up to 0.5s) depending on Settings.WaitScanRate.

The code snippet:

# timer start
if not myRegion.exists(myPattern, 0):
    #timer end

Now the result:

Settings.WaitScanRate = 10
Duration: 0.1s

Settings.WaitScanRate = 5
Duration: 0.2s

Settings.WaitScanRate = 4
Duration: 0.25s

Settings.WaitScanRate = 3
Duration: 0.33s (default setting... this is a significant delay)

Settings.WaitScanRate = 2
Duration: 0.5s

Settings.WaitScanRate = 1
Duration: 0.03s

Settings.WaitScanRate = 0.5
Duration: 0.03s

Settings.WaitScanRate = 0.1
Duration: 0.03s (seems the minimal one can reach on that operation)
...

__________

Bug location:
http://bazaar.launchpad.net/~raimund-hocke/sikuli/sikuli-api/view/head:/src/main/java/org/sikuli/script/Region.java#L1879

#L1881 int MaxTimePerScan = (int) (1000.0 / waitScanRate);
#L1882 int MaxTimePerScanSecs = MaxTimePerScan / 1000;
MaxTimePerScanSecs is equal to 0 when waitScanRate is strictly higher to 1, integer division rules in java...

#L1890 if (timeout < MaxTimePerScanSecs)
This is false (assuming timeout is 0), and additional delay ensue...

[RaiMan (raimund-hocke)]

LOL, absolutely right. Thanks for finding.

[RaiMan (raimund-hocke)]

This is my fix:

      int MaxTimePerScan = (int) (1000.0 / waitScanRate);
      int timeoutMilli = (int) timeout * 1000;
      long begin_t = (new Date()).getTime();
      do {
        long before_find = (new Date()).getTime();
        run();
        if (ifSuccessful()) {
          return true;
        } else if (timeoutMilli < MaxTimePerScan) {
          // instant return on first search failed if timeout very small or 0
          return false;
        }

[kurokrosk (kurokrosk)]

Looks good to me. I would have added parenthesis around timeout * 1000 to handle timeouts below 1s.

[RaiMan (raimund-hocke)]

yes, agreed.
thanks.