Comment by Paul Jereb on "For Speeds Sake, Stop Using APEX_JSON"

CommentFor Speeds Sake, Stop Using APEX_JSON

Paul Jereb

Sep 23, 2023

Hi Jon, thanks for your article!

After moving an implementation of generating JSON output for ORDS from APEX_JSON to JSON_OBJECT_T, we saw that:

generating the JSON itself became faster, but for some reason
"printing" it to ORDS is significantly slower using JSON_OBJECT_T than it is with APEX_JSON.

Out of interest, can you reproduce this behavior in your tests, or does adding "apex_util.prn ( l_json_clob, false );" take the same time, independent if the CLOB comes from JSON_OBJECT_T or APEX_JSON?

BR Paul

Jon Dixon

APEX, ORDS & the Oracle Database

Sep 23, 2023

Hi Paul. I haven't run this specific test, but I assume apex_util.prn takes the same amount of time regardless of what generated the CLOB. Could it be that JSON_OBJECT_T is generating a larger CLOB? Maybe try comparing the size of the CLOB using each method.

Jon Dixon yeah, I'd assume the same. The CLOBs are different, APEX_JSON seems to escape string values ("O'Brien" becomes "O\u0027Brien") resulting in a bigger CLOB, but faster print performance.

Escaping string values in JSON_OBJECT_T (using e.g. apex_escape.json) improves the print performance, but the serialization (to_clob) messes up the output by adding an additional backslash ("O'Brien" becomes "O\\u0027Brien").

Search Hashnode