A web site doc comment from user skong today points out a small issue
around the sample INSTR function given in plpgsql-porting.html that I
can't confirm (none of those dirty Oracle instances here today), but it
sounds legit.
A look at Oracle's documentation on the INSTR function at
http://docs.oracle.com/cd/B28359_01/olap.111/b28126/dml_functions_1103.htm
says that the 3rd input, position to start searching, cannot be zero.
skong says that Oracle will just return a 0 if you give it that invalid
input.
The INSTR implementation in the docs will instead search backwards from
the end of the string if you tell it to start at 0, same as if you gave
it a negative input. I think it's therefore possible to get the plpgsql
version to return a value in cases Oracle would instead return 0. Seems
like a straightforward thing to confirm and change the sample to do
differently; just have to add an explicit test for a 0 value of beg_index.