Convert zero-padded bytes to UTF-8 string

Question

I'm unpacking several structs that contain 's' type fields from C. The fields contain zero-padded UTF-8 strings handled by strncpy in the C code (note this function's vestigial behaviour). If I decode the bytes I get a unicode string with lots of NUL characters on the end.

>>> b'hiya\0\0\0'.decode('utf8')
'hiya\x00\x00\x00'

I was under the impression that trailing zero bytes were part of UTF-8 and would be dropped automatically.

What's the proper way to drop the zero bytes?

Theurer

Willing to accept an answer that can put comment to UTF-8's treatment of trailing bytes. – Theurer

My understanding is that the NULL codepoint in unicode should be encoded in utf-8 as a null byte, but because some languages use null to terminate a string there is an alternate encoding used e.g. in Java known as modified utf-8 which uses C0,80 to encode a null. See en.wikipedia.org/wiki/UTF-8#Modified_UTF-8 – Maria

you could fix it at the unpacking stage: if your input is always null-terminated then use ctypes.c_char_p type that converts C strings to Python bytes on input. See reading struct in python from created struct in c – Peugia

Answer 1

Either rstrip or replace will only work if the string is padded out to the end of the buffer with nulls. In practice the buffer may not have been initialised to null to begin with so you might get something like b'hiya\0x\0'.

If you know categorically 100% that the C code starts with a null initialised buffer and never never re-uses it, then you might find rstrip to be simpler, otherwise I'd go for the slightly messier but much safer:

>>> b'hiya\0x\0'.split(b'\0',1)[0]
b'hiya'

which treats the first null as a terminator.

Maria

I would suggest b'hiya\0x\0'.partition(b'\0')[0] instead. – Align

Good call. I always forget about partition. – Maria

Answer 2

Use str.rstrip() to remove the trailing NULs:

>>> 'hiya\0\0\0'.rstrip('\0')
'hiya'

Gadolinium

This might fail if the c-string has not been initialised with zeros or a old string has been overwritten by a shorter one. When a c-string-variable is changed, only the byte behind the last char is set to zero. – Windshield

That doesn't prove the output doesn't have trailing nulls, because if they're there, it doesn't get printed. However, if you use len(), then it'll probably help, as len doesnt care about trailing nulls. – Ney

Answer 3

Unlike the split/partition-solution this does not copy several strings and might be faster for long bytearrays.

data = b'hiya\0\0\0'
i = data.find(b'\x00')
if i == -1:
  return data
return data[:i]

Windshield

same as data[:data.find(0)] – Maryannemarybella

No, that code would fail if the buffer is full (contains no zeros). I.e. data = b'hiyafoo' – Windshield

Answer 4

I found this to be a neat solution:

''.join(chr(b) if b else '' for b in b'\0hello\0\0)

Skedaddle