Unexpected number of bytes (chunk size) in encoding?

[thorwhalen]

Noticed this unexpected behavior:

import recode
dh_codec = recode.mk_codec('dh')
assert len(dh_codec.encode([[1.0, 1]])) == 10

hd_codec = recode.mk_codec('hd')
assert len(hd_codec.encode([[1, 1.0]])) == 16

Which can be traced down to the builtin struct itself:

import struct
assert struct.calcsize('h') == 2
assert struct.calcsize('d') == 8
assert struct.calcsize('dh') == 10 == 2 + 8  # expected!

assert struct.calcsize('hd') == 16  # didn't expect this!

What's happening here?

How can I get what I expect? (10 in both cases.)

[charlie-dufort]

@sylvainbonnot explained it had to do with this optional prefix character ('@', '=', '<', '>', '!') which controls the byte order, size and alignment.
If the first character is not one of these, '@' is assumed.
Using '=' removes the "unexpected behavior":

>>> struct.calcsize('=hd') == struct.calcsize('=dh') == 10
True