trace_by_hash_readme

Link to original blog item:

My 10046 SQL Trace File Formatter

Notes:

• The awk script is run as a shell script, so script needs execute permissions
• This version runs on 10g and 11g (seperate 9i version with slight changes available)
• I used this script mostly on 10g and just made a couple of minor tweaks to also support 11g (does not convert cursor to sql_id yet)
• Cutting out smaller parts of a raw trace file with sqlhash_for_cursor=1 (ie cursor numbers replaced by hash_value) seems to cause tkprof to crash. I am sure it worked in past but may be wrong. Should still be able to profile small parts of trace file if with debug=1 and sqlhash_for_cursor=0)

“configure options” parameters (in the top of the script)

• debug
  • Set to 0 will switch off the extra PARSING IN CURSOR text on every EXEC
  • Set to 1 will switch on an extra PARSING IN CURSOR section on every EXEC
• sqlhash_for_cursor
  • Set to 0 switches off replacement of #cursor numbers with sql_hash_value
  • Set to 1 switches on replacement of #cursor numbers with sql_hash_value
• indent_level
  • Set to 0 will not indent the sql depth (dep=)
  • Set to any other value will indent output by n chars for each depth
• clock_time_on
  • Set to 0 switches off the approimate clock and delta time
  • Set to 1 switches on the approimate clock and delta time
• generate_hash_file
  • Set to 0 switches off generation of file containing sql statement dependency
  • Set to 1 switches on generation of file (hash.txt)
  • Note: requires a LOT of messing with the file, never fully implemented

Link to the script:

 trace_by_hash.awk