| 1 |
# @isaacs/testing-transitive-conflicted-peer |
| 2 |
|
| 3 |
This demonstrates a case where a transitive dependency of a project |
| 4 |
has a peer dependency which can only be placed in the root, but conflicts |
| 5 |
with the root's direct dependency. |
| 6 |
|
| 7 |
``` |
| 8 |
root -> (a, b@2) |
| 9 |
a -> (c), PEER(b@1||2) |
| 10 |
c -> PEER(b@1) |
| 11 |
``` |
| 12 |
|
| 13 |
The algorithm places `a` and `b@2` in the root, as it must. |
| 14 |
|
| 15 |
When seeking a placement for `c`, it hits a `CONFLICT` with the peerDeps |
| 16 |
`c->(b@1)` and `root->(b@2)`. Since that is not the deepestNestingTarget |
| 17 |
for `c`, it places it under `a` as the target. |
| 18 |
|
| 19 |
(NB: this is somewhat inefficient in this particular sort of situation, as |
| 20 |
it can lead to unnecessary duplication, and there is no practical benefit |
| 21 |
in this case to nesting `c` under `a`. But as a general heuristic, it |
| 22 |
prevents most conflicts, and the work to determine that nesting is |
| 23 |
effectively the same outcome as deduplicating can be expensive.) |
| 24 |
|
| 25 |
Then, resolving the `c->(b@1)` peer dep, it cannot place `b` under `a`, |
| 26 |
because `a` has a peerDep on `b`. So, the deepest nesting target for `b` |
| 27 |
is under `root`. |
| 28 |
|
| 29 |
However, there it hits a conflict, and checks to see if the `c` node (the |
| 30 |
nearest non-peer dependency leading to the conflicted node) is one of our |
| 31 |
project's direct dependencies. Since it is not, this is a warning, rather |
| 32 |
than an error. |