[PATCH] Performance improvements for getTileUrl

[wiktorn]

getTileUrl is called in EDT thread. Going through profiles in #16760 and #16734 I noticed that there are some easy optimizations for getTileUrl.

I've optimized getTileUrl for TMS and for WMS. Performance check was to call getTileUrl 10000 times in 20 rounds, each measured separately. From this 20 rounds I discarded 3 slowest results.

Effects for TMS:

x current
+ optimized
+--------------------------------------------------------------------------------+
|  + +                                                   xx                      |
| ++ +      + +                                          xx                      |
|+++++      +++    +        +                           xxxx  x       x  x      x|
||___M___A______|                                     |___M__A_____|             |
+--------------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  17           136           175           138     143.05882     11.294194
+  17            45            90            52     57.352941     12.338665
Difference at 95.0% confidence
        -85.7059 +/- 8.26402
        -59.9095% +/- 5.77666%
        (Student's t, pooled s = 11.828)

Effects for WMS:

x current
+ optimized
+--------------------------------------------------------------------------------+
|           +                                                                    |
|           +                                                 x  x               |
|          ++                                                 x  x               |
|   +      ++ +                                               xxxx               |
|+  +     +++++      +  +                                    xxxxxx  x x        x|
|     |_____A____|                                           |__MA____|          |
+--------------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x  17           201           226           204     206.17647     6.1464959
+  17           120           151           135     134.35294     7.4660999
Difference at 95.0% confidence
        -71.8235 +/- 4.77775
        -34.8359% +/- 2.31731%
        (Student's t, pooled s = 6.8382)

Few things I'd like to have reviewed:

• I put a tests for TemplatedTMSTileSource in JOSM repository, though the file comes from JMapViewer, but JMapViewer doesn't have any tests at all
• I'm not sure if all edge cases are covered in TemplatedTMSTileSorce

[Don-vip]

I've enabled tests in JMapViewer, they are now reported on Jenkins, see below:

• jenkins/job/JMapViewer/jdk=JDK8/lastCompletedBuild/testReport/
• jenkins/job/JMapViewer/jdk=JDK8/jacoco/

[Don-vip]

Also, if you add new tests making calls to real servers on the Internet, please flag them as integration tests (i.e add them in the FooTestIT class instead of FooTest). Unit tests must be fast and reliable, so cannot use external servers.

[wiktorn]

In 14399/josm:

GetTileUrl performance improvements for WMS

See: #16769

[Don-vip]

Is this ticket fixed?

[wiktorn]

Indeed it is.