Difference between revisions of "Postgres Dev."

From Earlham CS Department
Jump to navigation Jump to search
(Open Questions)
(Test Environment)
Line 32: Line 32:
 
I was able to locate the error messages generated by foreign key constraints. They come from one file, called ri_triggers.c
 
I was able to locate the error messages generated by foreign key constraints. They come from one file, called ri_triggers.c
  
I am currently creating a test enviroment where we can reproduce this error, and hopefully change the contents of the error message.
+
We have created a local test environment and accompanying script that fails on a fk constraint. When the SQL call fails, our modified error message gets printed with the namespace identification number in place of the relation name.
  
 
== Open Questions ==
 
== Open Questions ==

Revision as of 20:54, 20 April 2009

Postgres Internals Notes

Current Efforts

  • Declaration of RelationGetRelationNamespace:
postgresql-8.3.7/src/include/utils/rel.h
RelationGetRelationName(relation)
  • Prints out the name of the relation. This function is called directly in the printf of the error message in ri_triggers.c
RelationGetNamespace(relation)
  • There's a member variable: relnamespace in Form_pg_class data type.

It is of type id0, potentially representing a unique namespace identifier. I just replaced the appearance of RelationGetRelationName(relation) with RelationGetNamespace(relation) and am waiting on the test environment to see whether this function returns the id number of the namespace (schema) in our simple example, this should refer to the public namespace.


  • Error "violates foreign key constraint located in:
postgresql-8.3.7/src/backend/utils/adt/ri_triggers.c:

Test Environment

We are currently creating a test environment on db-devel.cs.earlham.edu where a query causes a foreign key constraint. I created a simpler case, with a single relation. When I query

SELECT * FROM (wrong relation name)

An error gets returned, and only the relation name is reported. Our fix plans to address this case, however searching for error messages containing "does not exist" yielded too many result to be useful.

I was able to locate the error messages generated by foreign key constraints. They come from one file, called ri_triggers.c

We have created a local test environment and accompanying script that fails on a fk constraint. When the SQL call fails, our modified error message gets printed with the namespace identification number in place of the relation name.

Open Questions

  • What's the difference between a tablespace and a namespace? - addressed
  • Is there a way to take a namespace identifier and get a namespace object?