Switching between case sensitive and case insensitive string comparison

Hi,

I have a B-Tree database which has the following three columns:
  + Integer field
  + Object name field
  + Data field

The integer field servers as the primary key, while the object name
field serves as the duplicated key. For example, I have records like
this:

  1 abc data
  1 def data
  ... ...
  2 abc data
  2 xyz data
  ... ...

Each row is uniquely identified with the integer ID and the object
name field.

For some integer keys, say 1 and 2, I want the object name field to be
case insensitive, while for some other integer keys, I want the object
name field to be case sensitive. How do I achieve this?

I call the "set_dup_compare" function to set up a callback function to
compare strings. The callback function has this signature:

int _dup_compare_fcn(DB* db, const DBT *arg1, const DBT *arg2)

One idea is to call the "set_dup_compare" function to switch function
pointer based on the integer key value. Another idea is to use the
"app_private" member in the DB structure to indicate case sensitivity.
Since the application is multi-threaded, I think that I need to mutex
a record lookup session to prevent the function pointer from being
switched.

But then I lose concurrency when populating the table: one thread can
populate records for an integer ID while another thread can populate
records for another integer ID.

Any suggestions?

Regards,
Jiebo Guan
NetIQ Corp