>>> On Wed, Sep 26, 2007 at 7:29 PM, in message <46FAF95D(dot)6070003(at)phlo(dot)org>,
"Florian G. Pflug" <fgp(at)phlo(dot)org> wrote:
> Kevin Grittner wrote:
>> I omitted the code I was originally considering to have it work against
>> files "in place" rather than as a filter. It seemed much simpler this
>> way, we didn't actually have a use case for the additional functionality,
>> and it seemed safer as a filter. Thoughts?
>
> A special "non-filter" mode could save some IO and diskspace by not actually
> writing all those zeros, but instead just seek to SizeOfWal-1 after writing
> the
> last valid byte, and writing one more zero. Of course, if you're gonna
> compress the WAL anyway, there is no point...
Right. And if you're not, why bother setting to zero? I couldn't invent
a plausible scenario where we would want to do the update in place, and
I'm afraid someone might be tempted to run it against "live" WAL files.
So I decided it was best to let it lie unless someone else had a real-
life situation where it was useful. Even then, I could write a bash
script to do it using the filter a lot faster than I could modify the C
code to safely deal with the files in-place, so I'm pretty skeptical.
-Kevin