aherbert commented on a change in pull request #111: TEXT-157: Remove rounding
from JaccardSimilarity and Distance
URL: https://github.com/apache/commons-text/pull/111#discussion_r263758068
##########
File path:
src/test/java/org/apache/commons/text/similarity/JaccardDistanceTest.java
##########
@@ -36,21 +36,23 @@ public static void setUp() {
@Test
public void testGettingJaccardDistance() {
- assertEquals(1.00d, classBeingTested.apply("", ""),
0.00000000000000000001d);
- assertEquals(1.00d, classBeingTested.apply("left", ""),
0.00000000000000000001d);
- assertEquals(1.00d, classBeingTested.apply("", "right"),
0.00000000000000000001d);
- assertEquals(0.25d, classBeingTested.apply("frog", "fog"),
0.00000000000000000001d);
- assertEquals(1.00d, classBeingTested.apply("fly", "ant"),
0.00000000000000000001d);
- assertEquals(0.78d, classBeingTested.apply("elephant", "hippo"),
0.00000000000000000001d);
- assertEquals(0.36d, classBeingTested.apply("ABC Corporation", "ABC
Corp"), 0.00000000000000000001d);
- assertEquals(0.24d, classBeingTested.apply("D N H Enterprises Inc", "D
& H Enterprises, Inc."),
- 0.00000000000000000001d);
- assertEquals(0.11d, classBeingTested.apply("My Gym Children's Fitness
Center", "My Gym. Childrens Fitness"),
- 0.00000000000000000001d);
- assertEquals(0.10d, classBeingTested.apply("PENNSYLVANIA",
"PENNCISYLVNIA"), 0.00000000000000000001d);
- assertEquals(0.87d, classBeingTested.apply("left", "right"),
0.00000000000000000001d);
- assertEquals(0.87d, classBeingTested.apply("leettteft", "ritttght"),
0.00000000000000000001d);
- assertEquals(0.0d, classBeingTested.apply("the same string", "the same
string"), 0.00000000000000000001d);
+ // Results generated using the python distance library using:
+ // distance.jaccard(seq1, seq2)
+ assertEquals(1.0, classBeingTested.apply("", ""));
+ assertEquals(1.0, classBeingTested.apply("left", ""));
+ assertEquals(1.0, classBeingTested.apply("", "right"));
+ assertEquals(0.25, classBeingTested.apply("frog", "fog"));
+ assertEquals(1.0, classBeingTested.apply("fly", "ant"));
+ assertEquals(0.7777777777777778, classBeingTested.apply("elephant",
"hippo"));
Review comment:
I understand where you are coming from. However it currently passes the test
because the underlying math is using `intersect / divide` with integers cast to
double. So the value can be exact and there is not really a way to do it with
less precision unless the algorithm is wrong.
I can update to using an epsilon of 1e-10 for all values that are not
expected to be 0.0 or 1.0. However that does make it possible someone reading
the test will think the output value is subject to imprecision because it is
not.
How about a comment in the test explaining where each value comes from, or
even the actual computation:
```
assertEquals(0.7777777777777778, classBeingTested.apply("elephant",
"hippo"));
// becomes
assertEquals(7.0 / 9.0, classBeingTested.apply("elephant", "hippo"));
```
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
