On 28.05.2015 15:01, Philip Martin wrote:
> Ivan Zhakov <i...@visualsvn.com> writes:
>
>> Also I don't understand what do you mean "messy file-based code"? Imho
>> code using svn_stream_write() that require pointer to length is more
>> messy.
> "messy" is not my word but the reason I prefer the stream code is that
> we have been moving towards it. Take this change:
>
> - SVN_ERR(svn_stream_printf(stream, pool, "%s\n", filename));
> + SVN_ERR(svn_io_file_write_full(file, filename, strlen(filename),
> + NULL, pool));
> + SVN_ERR(svn_io_file_putc('\n', file, pool));
>
> We have svn_stream_printf() and it makes the stream code neater. We
> could fix that by introducing svn_io_file_printf(). Either we add all
> the neat stream features to the file code or we attempt to move to the
> stream code.
Yes, and the former fits my definition of "messy". Even if we add a
number of svn_io functions to make file-based code as neat and readable
as stream-based code, we'd still be taking a step back in terms of
flexibility: you can stack and combine streams, but you can't do that
with file handles (other than by inventing a new file abstraction, which
is exactly what a stream is).
-- Brane
