Description

urswiss

opened on Jun 29, 2022

Issue body actions

After migrating from 1.4.2 to 1.5.2 a test failed. The result from BigDecimal to double is not the same anymore. Please have a look. Thanks.

https://github.com/urswiss/mapstruct-bigdecimal-to-double

Activity

1. Zegveld commented on Jun 29, 2022

   

   Zegveld

   on Jun 29, 2022

   Contributor

   More actions

   Here is the analysis:

   available method to use:

       default Long mapAmount(BigDecimal amount) {
           return amount.movePointRight(2).longValue();
       }

   Reason

   As it is legal to assign a long to a double, mapstruct will use this method because it is user defined.

   Generated code

   1.4.2

   simpleDestination.setRoundingDiff( source.getRoundingDifference().doubleValue() );

   1.5.2

   simpleDestination.setRoundingDiff( mapAmount( source.getRoundingDifference() ) );

   work around

   Use qualifiers to mark the mapAmount method as one specific for certain mappings and use qualfiedBy there to link them.

   What will we do about this

   I think this is also related to: #2840 and #1216
   Because if we pick the exact match, there might be a better workaround by manually creating the default double map(BigDecimal decimal) method.

   @filiphr: WDYT ?
2. filiphr

   added

   for:team-discussion

   on Jun 30, 2022
3. filiphr

   added this to the 1.6.0 milestone on Jun 30, 2022
4. filiphr commented on Jun 30, 2022

   

   filiphr

   on Jun 30, 2022

   Member

   More actions

   Thanks for the analysis @Zegveld. This is most definitely related to #2840 and #1216. We'll need to figure out what actually broke this behaviour.

   I think that we need to fix this inconsistency, because mapping from BigDecimal to double shouldn't use BigDecimal to long.

   Reacted by Urs Wiss
5. filiphr commented on Aug 21, 2022

   

   filiphr

   on Aug 21, 2022

   Member

   More actions

   After doing some analysis on this one and #2849 it seems like the work we did in #2320 broke this. I still need to analyse why this happened. This was not supposed to happen when we did that PR. What do you think @sjaakd? It feels like the fact that the return type can be assigned to the expected type is causing the problem.
6. filiphr

   added

   backport-1.5.xIssues that should be backported to 1.5Issues that should be backported to 1.5

   on Aug 21, 2022
7. filiphr

   added

   bug

   and removed

   for:team-discussion

   on Sep 26, 2022
8. filiphr

   added a commit that references this issue on Sep 26, 2022

   mapstruct#2840, mapstruct#2913, mapstruct#2921: MethodMatcher should …

   50db7bd
9. filiphr commented on Sep 26, 2022

   

   filiphr

   on Sep 26, 2022

   Member

   More actions

   I am happy to announce that we have managed to find the root cause for this problem and we have a pending fix in #3032

   Reacted by Silvio Wangler and Urs Wiss
10. filiphr

    self-assigned this

    on Sep 26, 2022
11. filiphr

    modified the milestones: 1.6.0, 1.5.3 on Sep 26, 2022
12. filiphr

    closed this as completedin #3032on Sep 27, 2022
13. filiphr

    added 2 commits that reference this issue on Sep 27, 2022

    #2840, #2913, #2921: MethodMatcher should not match widening methods

    73e8fd6

    #2840, #2913, #2921: MethodMatcher should not match widening methods

    73f70b1

Issue actions