You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This can be problematic in other parts of Epsilon such as ETL's default transformation strategy which assume that model elements have unique hash codes. Options to explore:
Add an ordinal nubmer field to CSV model elements to avoid duplicate hash codes
Change from hash codes to system identities in ETL's FastTransformationStrategy
The text was updated successfully, but these errors were encountered:
System identifiers can be problematic with Ecore elements (Epackage,
EClass, etc) as sometimes the same metamodel is loaded twice, so the same
EClass has two different system identifiers (what Ed Willink nicely called
metamodel schizophrenia)
On Thu, May 1, 2025, 5:46 AM Dimitris Kolovos ***@***.***> wrote:
*kolovos* created an issue (eclipse-epsilon/epsilon#180)
<#180>
This can be problematic in other parts of Epsilon such as ETL's default
transformation strategy which assume that model elements have unique hash
codes. Options to explore:
- Add an ordinal nubmer field to CSV model elements to avoid duplicate
hash codes
- Change from hash codes to system identities in ETL's
FastTransformationStrategy
—
Reply to this email directly, view it on GitHub
<#180>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAQOU3L76EHVFKB7N6V3VPL24ICS7AVCNFSM6AAAAAB4HXHHYOVHI2DSMVQWIX3LMV43ASLTON2WKOZTGAZTGNRZGE2DEMY>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
I'd avoid making a broad change to ETL as it may have unintended consequences. It may be better to change CSV rows so they have different hashcodes for each other.
When I tried adding a row number to a CSV row, I realized it'd be harder than I expected to keep it up to date as rows are removed/inserted in the middle of the file. We wouldn't really want to expose such a pseudo-row number to users, as its behaviour may not be very reliable.
Why not change the internal representation of a row to a LinkedHashMap subclass which reverts hashCode+equals to be based on object identity?
This can be problematic in other parts of Epsilon such as ETL's default transformation strategy which assume that model elements have unique hash codes. Options to explore:
FastTransformationStrategy
The text was updated successfully, but these errors were encountered: