| 1 |
A variation on 25. |
| 2 |
|
| 3 |
This test justifies tracking dependencies within winning sets. |
| 4 |
|
| 5 |
In 25, line 4 is the only winner for y, and so it is required to |
| 6 |
survive. Then, line 4 requires line 1, and line 1 requires line |
| 7 |
6. Because line 6 is required, line 2 (with which line 6 is a tie), |
| 8 |
can be removed. |
| 9 |
|
| 10 |
But here, we have an additional tie, this time for y, between line 4 |
| 11 |
and line 7, and as such we have no "early" requirement that line 1 |
| 12 |
(the comprehension) survives. So we have a tie for x between lines 2 |
| 13 |
and 6, and a tie for y between lines 4 and 7. The tie for x gets |
| 14 |
solved first. |
| 15 |
|
| 16 |
If we did not track dependencies *within* a winning set, then the |
| 17 |
following would occur: |
| 18 |
1. we would choose line 2 to win (based on file position) |
| 19 |
2. line 2 requires line 1 |
| 20 |
3. line 1 requires line 6. |
| 21 |
So we would end up with both line 2 and line 6 surviving. |
| 22 |
|
| 23 |
But because we identify dependencies between winning nodes *within* |
| 24 |
the same set, we find than line 2 transitively depends on line 6, and |
| 25 |
as such, when choose a winner, we make sure we choose line 6. |
| 26 |
|
| 27 |
Having chosen the winner for x, we then chose the winner for y between |
| 28 |
lines 4 and 7. This is much simpler - we have a tie of 1-1 for the |
| 29 |
seenCounts, and so we just pick line 4 because 4 < 7. |
| 30 |
|
| 31 |
-- a.cue -- |
| 32 |
if x == 5 { |
| 33 |
x: 5 |
| 34 |
x: int |
| 35 |
y: "hi" |
| 36 |
} |
| 37 |
x: 5 |
| 38 |
y: "hi" |
| 39 |
-- out/trim-v3 -- |
| 40 |
== a.cue |
| 41 |
if x == 5 { |
| 42 |
y: "hi" |
| 43 |
} |
| 44 |
x: 5 |
| 45 |
-- out/trim-v3-noshare -- |
| 46 |
== a.cue |
| 47 |
if x == 5 { |
| 48 |
y: "hi" |
| 49 |
} |
| 50 |
x: 5 |