When does python compile the constant string letters, to combine the strings into a single constant string?

Question

such as

In [9]: dis.disassemble(compile("s = '123' + '456'", "<execfile>", "exec"))
  1           0 LOAD_CONST               3 ('123456')
              3 STORE_NAME               0 (s)
              6 LOAD_CONST               2 (None)
              9 RETURN_VALUE

I want to know, when does python combine the constant string as the CONST. If possible, please tell me which source code about this at cpython(whatever 2.x, 3.x).

Isn't that your answer right there? What are you confused about? — Joel Cornett, Jan 24 '13 at 03:15
I think OP wants to know where '123' + '456' turned into '123456' — wim, Jan 24 '13 at 03:18
@JoelCornett: `'123 + 456'` evaluates in exactly the same way as does `'123456'`. There seems to be no differentiation between `'123`' and `'456'` in `'123 + 456'` — inspectorG4dget, Jan 24 '13 at 03:19

score 11 · Accepted Answer · edited Nov 08 '20 at 00:50

11

It happens whenever the combined string is 20 characters or fewer.

The optimization occurs in the peephole optimizer. See line 219 in the fold_binops_on_constants() function in Python/peephole.c: http://hg.python.org/cpython/file/cd87afe18ff8/Python/peephole.c#l149

edited Nov 08 '20 at 00:50

codeforester

39,467
16
112
140

answered Jan 24 '13 at 03:19

Raymond Hettinger

216,523
63
388
485

6

I wrote Python's peephole optimizer :-) – Raymond Hettinger Jan 24 '13 at 03:22
1

Investiganting for a [probably duplicate of this](https://stackoverflow.com/questions/49427151/how-does-python-determine-if-two-strings-are-identical/) I found out that folding has been recently moved to [`ast_opt.c`](https://github.com/python/cpython/blob/master/Python/ast_opt.c), and the maximum size for folded strings is now defined as 4096 (`MAX_STR_SIZE`). – jdehesa Mar 22 '18 at 12:07

score 4 · Answer 2 · edited May 23 '17 at 12:14

@Raymond Hetting's answer is great, vote for that (I did). I'd make this a comment, but you can't format code in a comment.

If you go over the 20 character limit, the disassembly looks like:

>>> dis.disassemble(compile("s = '1234567890' + '09876543210'", "<execfile>", "exec"))
  1  0 LOAD_CONST  0 ('1234567890')
     3 LOAD_CONST  1 ('09876543210')
     6 BINARY_ADD
     7 STORE_NAME  0 (s)

But in the case where you have two string literals, remember you can leave out the + and use String literal concatenation to avoid the BINARY_ADD (even when the combined string length is greater than 20):

>>> dis.disassemble(compile("s = '1234567890' '09876543210'", "<execfile>", "exec"))
  1  0 LOAD_CONST  0 ('123456789009876543210')
     3 STORE_NAME  0 (s)

When does python compile the constant string letters, to combine the strings into a single constant string?

2 Answers2

Linked