I could add content negotiation, but that'd be redoing some stuff I did for the DS stats rewrite. Plus, I agree that this endpoint shouldn't exist. I'll make it use the custom format, but it'll be harder to tell why when tests fail.
On Fri, Sep 13, 2019 at 12:58 PM Rawlin Peters <[email protected]> wrote: > We have tools that use the /dbdump API for things like DB backups and > DB testing. If the API output changed out beneath those things, they > would 100% break. IMO the default output should remain as it is today, > and if we think a new format is needed then we should just add an > optional query parameter to the /dbdump endpoint. For example: > /api/1.1/dbdump (uses default compressed format) > /api/1.1/dbdump?format=compressed (same as default) > /api/1.1/dbdump?format=sql (uses new "SQL script" output format for > whatever reason) > > That is, if we think we even have a valid use case for a new format. > We should be careful not to add things to the API that we don't really > need. Already it seems like this endpoint is very dangerous and > _probably_ shouldn't exist (although it does and things depend on it > now). If a test can just read the current compressed format and > convert it to a sql script for analysis or whatever, that's probably > better than adding support for various output formats. > > - Rawlin > > On Fri, Sep 13, 2019 at 10:59 AM ocket 8888 <[email protected]> wrote: > > > > AFAIK the current output format is incomprehensible to anything but > > pg_restore > > > > On Fri, Sep 13, 2019 at 10:21 AM Jeremy Mitchell <[email protected]> > > wrote: > > > > > If pg_restore can handle both (the current binary and your suggested > plain > > > sql) seamlessly then doesn't seem like a problem to me to change the > format > > > to plain SQL but i can't speak for everyone using this endpoint. is > > > everyone using the endpoint's response in tandem with pg_restore? > > > > > > if it helps testability, i think your suggestion makes a lot of sense. > > > > > > jeremy > > > > > > On Thu, Sep 12, 2019 at 11:26 AM ocket 8888 <[email protected]> > wrote: > > > > > > > I'm rewriting the dbdump endpoint (PR #3912) and the output format > used > > > by > > > > Perl is a binary, compressed SQL script. It makes editing and even > > > testing > > > > just a bit harder than it could be. The data is still compressed for > > > > transfers, but a plain SQL script is much easier for a human to > analyze. > > > > pg_restore can work with any format that pg_dump outputs - and it > > > > automatically detects which one to use, so the user doesn't even > need to > > > > know, really. > > > > > > > > But is this a problem for anyone, for any reason? > > > > > > > >
