On Tue, May 27, 2014 at 12:44 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Amit Langote <amitlangote09@gmail.com> writes:
>> In a custom pgbench script, it seems convenient to be able to split a
>> really long query to span multiple lines using an escape character
>> (bash-style). Attached adds that capability to read_line_from_file()
>> in pgbench.c
>
> This seems pretty likely to break existing scripts that happen to contain
> backslashes. Is it really worth the compatibility risk?
>
> The patch as written has got serious problems even discounting any
> compatibility risk: it will be fooled by a backslash near the end of a
> bufferload that doesn't end with a newline, and it doesn't allow for
> DOS-style newlines (\r\n), and it indexes off the array if the buffer
> contains *only* a newline (and, assuming that it fails to crash in that
> case, it'd also fail to note a backslash that had been in the previous
> bufferload).
>
Sorry, the patch was in a really bad shape. Should have pondered these
points before submitting it.
Even if I drop the backslash line-continuation idea and decide to use
semi-colons as SQL command separators, given the compatibility issues
mentioned downthread, it would not be worthwhile.
--
Amit