OpenJPA
  1. OpenJPA
  2. OPENJPA-576

Recursively embedded entities generate an extra and reduntant SQL call to the database.

    Details

    • Type: Bug Bug
    • Status: Closed
    • Priority: Major Major
    • Resolution: Fixed
    • Affects Version/s: None
    • Fix Version/s: 1.1.0
    • Component/s: jdbc
    • Labels:
      None
    • Environment:

      Description

      Consider the following persistent class model. The accessor methods are removed for the sake of brevity.

      public class EmbeddedOwnerPC {

      private int id1;
      private int id2;
      private String stringField;
      private EmbeddedPC embedded;
      private ComplexEmbeddedPC complexEmbedded;
      }

      public class EmbeddedPC {
      private int intField;
      private String stringField;
      }

      public class ComplexEmbeddedPC extends RecursivelyEmbeddedPC {

      private String stringField;
      private EmbeddedOwnerPC ownerField;
      private Set stringSet = new HashSet();
      }

      public class RecursivelyEmbeddedPC {

      private EmbeddedPC embedded;
      }

      When trying to retrieve the EmbeddedOwner entity with a given fetch plan that includes the complexEmbedded and the embedded fields in the RecursivelyEmbeddedPC class it generate the following 2 SQL statements to load the entity.

      SELECT t0.TYP, t0.VERSN, t0.EMBEDDED_NULL, t0.ID1, t0.ID2, t0.INTFIELD1,
      t0.STRINGFIELD3, t0.STRINGFIELD FROM EMBEDDEDOWNERPC t0 WHERE t0.ID1 = ? AND t0.ID2 = ? FOR UPDATE

      SELECT t0.ID1, t0.ID2 FROM EMBEDDEDOWNERPC t0 WHERE t0.ID1 = ? AND t0.ID2 = ?
      FOR UPDATE

      The second call is reduntant and not needed. The fix is to the EmbedFieldStrategy class load method to ensure that all fields that are required to be loaded as per the fetch plan and included from the result set and ONLY when this is not true then generate the extra call otherwise return a hollow embedded entity. In the above test case the extra call was generated to get the stringSet field which was not specified in the fetch plan but the call is made out nevertheless which just selects the primary keys with the same where clause.

      1. OPENJPA-576.patch
        2 kB
        Sandeep Shrivastava
      2. OPENJPA-576-v2.patch
        3 kB
        Sandeep Shrivastava
      3. OPENJPA-576-v2.patch
        2 kB
        Sandeep Shrivastava

        Activity

        Hide
        Sandeep Shrivastava added a comment -

        Suggested fix for the issue.

        Show
        Sandeep Shrivastava added a comment - Suggested fix for the issue.
        Hide
        Patrick Linskey added a comment -
        • // after loading everything from result, load the rest of the
        • // configured fields
        • if (loaded)
        • em.load(fetch);
          + // After loading everything from result, load the rest of the
          + // configured fields if anything is missing.
          + if (!loaded) { + em.load(fetch); + }

        This seems like a relatively significant semantic change. Can you describe why you inverted the initialization value of |loaded| and why you changed this logic at the end?

        Also, what tests have you run this fix against?

        Show
        Patrick Linskey added a comment - // after loading everything from result, load the rest of the // configured fields if (loaded) em.load(fetch); + // After loading everything from result, load the rest of the + // configured fields if anything is missing. + if (!loaded) { + em.load(fetch); + } This seems like a relatively significant semantic change. Can you describe why you inverted the initialization value of |loaded| and why you changed this logic at the end? Also, what tests have you run this fix against?
        Hide
        Patrick Linskey added a comment -

        From a conversation with Sandeep:

        > We initialize the boolean loaded = true; so we go with the assumption
        > that we don't need to reload anything and everything we need is already
        > in the result set. The variable gets set to false only if a required
        > field is not loaded as per the fetch configuration and then we retry
        >
        > if (!loaded)

        { > em.load(fetch); > }

        >
        > So the embedded entity remains hollow if some of the fields are not
        > present in the result set and they are not present in the fetch
        > configuration.

        Show
        Patrick Linskey added a comment - From a conversation with Sandeep: > We initialize the boolean loaded = true; so we go with the assumption > that we don't need to reload anything and everything we need is already > in the result set. The variable gets set to false only if a required > field is not loaded as per the fetch configuration and then we retry > > if (!loaded) { > em.load(fetch); > } > > So the embedded entity remains hollow if some of the fields are not > present in the result set and they are not present in the fetch > configuration.
        Hide
        Sandeep Shrivastava added a comment -

        Revised patch based on comments from Abe.

        Show
        Sandeep Shrivastava added a comment - Revised patch based on comments from Abe.
        Hide
        Sandeep Shrivastava added a comment -

        Further refinement of the patch based on feedback from Abe White.

        Show
        Sandeep Shrivastava added a comment - Further refinement of the patch based on feedback from Abe White.

          People

          • Assignee:
            Unassigned
            Reporter:
            Sandeep Shrivastava
          • Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

            • Created:
              Updated:
              Resolved:

              Time Tracking

              Estimated:
              Original Estimate - 24h
              24h
              Remaining:
              Remaining Estimate - 24h
              24h
              Logged:
              Time Spent - Not Specified
              Not Specified

                Development