Consider the schema:

CREATETABLE `Account` (`Id`bigint(20) unsignedNOT NULL AUTO_INCREMENT,`Relation`bigint(20) unsigned DEFAULTNULL,`Company`varchar(50) DEFAULTNULL,`MainAddressId`bigint(20) unsignednot null,`AlternateAddressId`bigint(20) defaultnull,`CreateDate` datetime DEFAULTCURRENT_TIMESTAMP,`ModifyDate` datetime DEFAULTCURRENT_TIMESTAMPONUPDATECURRENT_TIMESTAMP,PRIMARY KEY (`Id`)) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;CREATETABLE `Locations` (`Id`bigint(20) unsignedNOT NULL AUTO_INCREMENT,`Address`VARCHAR(200)NOT NULL,`City`VARCHAR(80) defaultnull,`CreateDate` datetime DEFAULTCURRENT_TIMESTAMP,`ModifyDate` datetime DEFAULTCURRENT_TIMESTAMPONUPDATECURRENT_TIMESTAMP,PRIMARY KEY (`Id`));

and queries

-- name: FindAccountById :oneSELECTsqlc.embed(a),sqlc.embed(locMain),sqlc.embed(c)FROM Account aLEFT JOIN Locations locMainONa.MainAddressId=locMain.IdLEFT JOIN Locations locAltONa.AlternateAddressId=locAlt.IdWHEREa.Id= ?LIMIT1;

when generating the code (mysql driver), the generated methodFindAccountById will be returning a customFindAccountByIdRow where the Id fields coming from different tables are named likeId,Id_2 andId_3.

I'm not sure if there's something that can be currently done to easily differentiate or have the generated code use table aliases to tell the tables apart (so that the resulting struct elements would be calledAdMain_Id andAdAlt_Id respectively . Is there something that can be done to easily manage this situation? Could this be an improvement? Is this something that's more related to the mysql driver than sqlc?

EDIT:

rename:spotify_url:"SpotifyURL"

doesn't seem useful in this case. The name of the resulting struct field depends on context (eg: every table may have an ID, but whether it ends up being Id_2 or Id_3 depends on the exact query and what joins happen).

EDIT 2:
embeds don't seem to help either, since it would be needed to be able to embed by alias as opposed to table name when doing a join. Something likesqlc.embed(adAlt) isn't recognised as per example above. I guess mainly the issue is with multiple joins on the same table.

EDIT 3:
Some playground examples:

https://play.sqlc.dev/p/37c12941f7f1943997f4a628b82fb6be33d852a71614bb581a3ee14d71c280eb - initial version, fields mangled

https://play.sqlc.dev/p/8aa8bb4478ca2f7a54533ec2d66a167e580a93f6879ab6af4c5ead6e6b6e989e - trying embed, but can't resolve

https://play.sqlc.dev/p/31f018cd6e834cd2ccd9122789cd91c8adf29e7ea5e478e6a29dfafd2846903e - this is weird, it seems to resolve but the resulting embeds are numbered, so rendering them still confusing (as one would have to remember the order of the joins)

Questions:

• could/should the embeds also resolve table aliases that are longer than one single letter?
• if so, could the alias be used to identify fields/embeds instead of numbers ?