SQLx Next (0.7+) Tracking Issue

[mehcode]

Things are taking longer than we'd like on this so I'm putting up a tracking issue to share some insight on what remains to be done in the next branch.

• Type-safe macros ( query!, etc. )

• Transactions

• Streams (fetch)

• Pool

     PgConnectOptions::parse("postgres://....")?
       .tls( .... )
       .pooled() // -> PgPoolOptions
       .max_connections(10)
       .connect().await? // Pool (either this or finish)
       .finish() // Pool (either this or connect) 

• Database Errors

• Generalized Query Placeholders [Next] [Feature] Generalized query placeholders (including support for expanding arrays into multiple placeholders) #875

  • Positional {0} becomes $n or ?n or @pn, etc.
  • Named {name} is resolved to a positional parameter so this works anywhere - not just sqlite
  • Zero or More Repetition {name+} = NULL or $1, $2, ...
  • Field Expressions in parens {(x.y.z)}
  • Implicit parameter scope
  • Explicit parameter scope with ..some_struct Struct update rest pattern syntax support for named bindings #591

• Runtime named parameters

• TLS

• Any (Driver)

• Derives

  • FromRow
  • Type

• Query logging

• Statement caching

• MySQL

  • Type parity
  • Support for MySQL 5.5
  • Use Error::Client for client errors

• PostgreSQL

  • Support for SASL (authentication)
  • Type parity
  • Listener
  • Database error

• SQLite

The PostgreSQL and MySQL drivers are done enough to test and prod the basics if anyone has time to poke it.

[link2xt]

Does the next branch include all the fixes from the master branch? Are fixes cherry-picked, merged or is the branch occasionally rebased? Currently it says the branch is 1850 commits behind master.

[mehcode]

The next branch is a full rewrite of at least the core that is being developed in parallel at the moment. Postgres and MySQL are nearing feature parity with master. Once that happens we'll port the test suite and then audit the comparison to ensure we don't miss anything important. I know there will probably be some bumps but there are some good reasons for why a lot of these changes couldn't have easily been done iteratively, notably the runtimes and cancellation safety.

[gyzerok]

@mehcode how up to date are checkboxes? Would you mind updating them?

[amitavaghosh1]

Could you guys also please look at the enum issue in mysql? where query_as! works fine but query_as() fails to do it properly. Right now we are overriding FromRow manually

[nyurik]

Would it be possible to re-release the old 0.6 as 0.7 with just the breaking libsqlite3-sys bump to 0.25.2?

All pending changes in the main would then go to 0.8 instead. It's not like numbers have much "value" in of itself, easy to increment - and minor bump would indicate the breaking nature of the change, even though it would contain just a tiny but breaking fix.

[galah92]

Would it be possible to re-release the old 0.6 as 0.7 with just the breaking libsqlite3-sys bump to 0.25.2?

+1

[lu-zero]

If somebody volunteers time to prepare such 0.6+sqlite-fix branch could it be published once ready?

[nyurik]

I created a new PR that just bumps the sqlite3 #2455 based on the v0.6.3 release. See the description there, as it requires a new branch in order to be merged.

[mrl5]

hello, I noticed that changelog doesn't mention breaking change related to sqlx prepare - namely --merged flag is no longer supported.

to reproduce

cargo install --version 0.7.0 sqlx-cli

and then run

cargo sqlx prepare --merged -D "${DB_URI}"

actual result

error: Found argument '--merged' which wasn't expected, or isn't valid in this context

	If you tried to supply `--merged` as a value rather than a flag, use `-- --merged`

USAGE:
    cargo sqlx prepare [OPTIONS] --database-url <DATABASE_URL> [-- <ARGS>...]

For more information try --help

expected result
I'm not really sure. Does it mean that with 0.7.0 I no longer need to have sqlx-data.json?

[stefankreutz]

Hey This change is described in the breaking changes section at the top of the changelog. The actual CLI flag `--merged` isn't mentioned in the changelog though.

The offline mode for the queries has been changed to use a separate file per query!() invocation, which is intended to reduce the number of conflicts when merging branches in a project that both modified queries. See [#2363] for details.

Regards Stefan

…

On Sun, Jul 09, 2023 at 04:08:11AM -0700, Jakub Kołodziejczak wrote: hello, I noticed that changelog doesn't mention breaking change related to `sqlx prepare` - namely `--merged` flag is no longer supported. **to reproduce** ```console cargo install --version 0.7.0 sqlx-cli ``` and then run ```console cargo sqlx prepare --merged -D "${DB_URI}" ``` **actual result** ``` error: Found argument '--merged' which wasn't expected, or isn't valid in this context If you tried to supply `--merged` as a value rather than a flag, use `-- --merged` USAGE: cargo sqlx prepare [OPTIONS] --database-url <DATABASE_URL> [-- <ARGS>...] For more information try --help ``` **expected result** I'm not really sure. Does it mean that with `0.7.0` I no longer need to have `sqlx-data.json`? -- Reply to this email directly or view it on GitHub: #1163 (comment) You are receiving this because you are subscribed to this thread. Message ID: ***@***.***>

[mrl5]

unfortunately after migrating to 0.7.0 and changing --merged to --workspace here

I'm getting:

warning: no queries found

and empty .sqlx directory on top level - I've tried to audit https://github.com/launchbadge/sqlx/blob/v0.7.0/sqlx-cli/src/prepare.rs but I have no clue why it's unable to find queries :/

update

ok so I had to:

1. bump sqlx version also in Cargo.toml - seems like sqlx-cli requires version compatibility
2. remove offline feature:

-sqlx = { version = "^0", features = [
+sqlx = { version = "0.7.0", features = [
     "chrono",
     "macros",
     "migrate",
-    "offline",
     "postgres",

[mrl5]

seems like #2086 introduced breaking change also for projects that don't do custom impl of PgHasArrayType (contrary to #2086 (comment))

more info in #2611