Sounds like a new myth to me. I've never heard of such a thing.

Besides, a 1,000 line package might compile to be less p-code then a 100 line package.

The fact is -- we page in code in 4k'ish chunks, regardless of size. Once upon a time, many many versions ago, we paged in entire packages/procedures -- for many years now, that has not been the case.

So, it is a myth with a kernel of (old) truth behind it -- but an "unachievable" myth. Were the developers supposed to sit there and add/remove code until the compiled code_size was "right on"??? Impossible, totally impossible.

Totally "non-sensible" and hopefully an almost unknown 'technique'

Reminds me of my Ada programming days. Someone tried to make me ALPHABETIZE my subroutines -- declare them in order alphabetically, define them in order -- alphabetically. They didn't like to read code on "screen", they wanted to print it out to read it and it would help them find stuff (since they couldn't "search" the paper). The end result - I wrote a script to rename all of my routines to:

aa_some_function;
ab_another_function;
ac_......

and so on (and used a big "pragma renames" at the bottom so I didn't have to hack the rest of the code, sort of like a synonym for a procedure). Met the letter of the law, but didn't do anything for anyone.

Here is a query you can use to see that this is totally futile:


1 select trunc(source_size/1024) src_size_k,
2 min(code_size),
3 max(code_size),
4 max(code_size)-min(code_size) diff,
5 count(code_size) cnt
6 from dba_object_size
7 where source_size > 0
8 and type in ( 'PROCEDURE', 'FUNCTION', 'PACKAGE BODY' )
9 group by trunc(source_size/1024)
10 having max(code_size)-min(code_size) > 0
11* order by 4
ops$tkyte@ORA817DEV.US.ORACLE.COM> /

SRC_SIZE_K MIN(CODE_SIZE) MAX(CODE_SIZE) DIFF CNT
---------- -------------- -------------- ---------- ----------
68 16006 16492 486 2
21 6187 6722 535 3
24 6120 6738 618 2
12 1867 2627 760 3
28 5684 6994 1310 2
74 17689 19007 1318 3
20 4954 6555 1601 2
26 4135 6074 1939 3
18 3170 5187 2017 3
150 30077 32203 2126 2
0 149 2383 2234 65
1 0 2239 2239 19
67 15850 18342 2492 2
9 1450 4582 3132 5
10 3191 6597 3406 3
31 6554 9987 3433 3
16 1469 5034 3565 7
27 5006 8935 3929 2
2 331 4375 4044 9
4 268 4549 4281 4
3 1502 5919 4417 2
132 34193 38985 4792 2
5 594 5520 4926 8
122 25256 31011 5755 2
6 1604 8053 6449 5
47 6113 12649 6536 2
11 2736 9588 6852 3
7 1135 9943 8808 4
15 4680 16379 11699 2
13 3321 15039 11718 4
22 3289 16350 13061 7
32 21828 35886 14058 2
54 15473 30590 15117 2
8 2671 17999 15328 5
25 4652 21423 16771 3
23 5379 22723 17344 3
38 8972 26923 17951 2
34 6352 24784 18432 2
17 3951 24080 20129 3
45 10423 44512 34089 3
71 8215 57543 49328 3

41 rows selected.

So, for example, there is some code out there that is 71k in size. The smallest results in about 8k of p-code, the largest in about 57k. Hmmm, what would they do then???