[chore][graph] Remodel node id as attribute sets #11344
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #11344 +/- ##
==========================================
+ Coverage 92.07% 92.08% +0.01%
==========================================
Files 432 432
Lines 20299 20339 +40
==========================================
+ Hits 18690 18730 +40
Misses 1249 1249
Partials 360 360 ☔ View full report in Codecov by Sentry. |
djaglowski
force-pushed
the
graph-attributes
branch
3 times, most recently
from
f3779d2
to
b299e68
Compare
djaglowski
force-pushed
the
graph-attributes
branch
from
d251016
to
f05591b
Compare
| `connector "nop/conn" (traces to traces) -> ` + | ||
| `processor "nop" in pipeline "traces/1" -> ` + | ||
| `connector "nop/conn1" (traces to traces)`, | ||
| `connector "nop/conn1" (traces to traces) -> ` + |
There was a problem hiding this comment.
Why the change in test here?
There was a problem hiding this comment.
It's really just a change in where the start of the cycle is reported, but it still reports the same cycle deterministically.
What it comes down to is that this PR changes the way node IDs are generated. Previously, we built a particualr string for each kind of component. Now we build the ID from the attribute set. Both methods are deterministic, but the graph presumably contains some logic where the node ids are considered in an order that is sensitive to their values (e.g. sort the ids, then for each one, follow the edges, etc). The logical graph that is produced is the same, but when a cycle is encountered, the order in which it is reported is affected by the order in which nodes were considered.
|
This PR was marked stale due to lack of activity. It will be closed in 14 days. |
This PR formalizes the notion that attributes are the defining identifiers for component instances in the graph package. It moves us towards the proposal described in #11343.