I would like to revisit this thread briefly. I believe Jerry was referring back to Lex's response: http://www.freelists.org/archives/oracle-l/09-2004/msg00243.html
He also mentioned hexstr from AskTom: http://asktom.oracle.com/pls/ask/f?p=4950:8:5671668926260613952::NO::F4950_P8_DISPLAYID,F4950_P8_CRITERIA:707586567563
Woflgang speculated that Oracle only encodes the first 6 characters, and will populate actual_value for character fields "when it needs to".
Given all this, it seems that there is no way to reverse-engineer a bucket on a character column consistently. Is that still true in 10gR2? In my own testing, this seems to be the case, especially if the column is larger than 6 characters. Even when it is not (when the column is less than 6 characters), the "encoding" process still makes decoding rather difficult. I used the hexstr function but it is too easy to get control characters in the stream (even using substr).
Is there any other way to analyze the bucket information for histograms on character fields? Thanks,
-- Charles Schultz