use sqitch in existing deployments

[jmls]

So, I've been manually using pgMigrate and have come across sqithc. Loving the idea of proper version control :)

However, as this is an existing app, how do I init sqitch with the existing schema ? I do not want )(obviously) to have to create manual deploy and revert scripts for every table, function etc

To add a slight twist , there are a couple of deployments that are not in sync with the latest dev version - how would I bring these into sync with development if I switch to sqitch ?

[theory]

Hi @jmls. There's actually a FAQ for that!

For your second question, you might consider starting with a dump of your production database for the first change, and then create a second with just the current changes in dev. How you diff those depends on your current tooling, though. Hopefully it's not too many changes.

I have, in the past, put in the effort of turning a database dump into individual changes. Usually took a couple days, but it all depends on the complexity. IIRC, a while back @adunstan started work on a script to parse a schema-only pg_dump file and turn it into changes, but I don't know if it got beyond the experimental phase.

HTH,

David

[adunstan]

On 2024-09-18 We 2:29 PM, David E. Wheeler wrote: Hi @jmls <https://github.com/jmls>. There's actually a FAQ for that <https://github.com/sqitchers/sqitch/wiki/How-do-I-get-started-with-an-existing-database%3F>! For your second question, you might consider starting with a dump of your production database for the first change, and then create a second with just the current changes in dev. How you diff those depends on your current tooling, though. Hopefully it's not too many changes. I have, in the past, put in the effort of turning a database dump into individual changes. Usually took a couple days, but it all depends on the complexity. IIRC, a while back @adunstan <https://github.com/adunstan> started work on a script to parse a schema-only |pg_dump| <https://www.postgresql.org/docs/current/app-pgdump.html> file and turn it into changes, but I don't know if it got beyond the experimental phase.

Umm, nope, don't think so. Message ID: ***@***.***>

[theory]

On Sep 18, 2024, at 14:50, Andrew Dunstan ***@***.***> wrote: Umm, nope, don't think so.

Really? I could have sworn. Would have been around 10 years ago though.