j0sh / ocaml-gensqlite Goto Github PK

Adding a sigil or some other built-in support to return the last inserted rowid would avoid having to carry the DB handle around in the application alongside gensqlite's generated query functions.

Some suggestions here: https://www.reddit.com/r/ocaml/comments/3gunm4/ocamlgensqlite_ocaml_ppx_preprocessor_to_generate/cu1yr0g

Hi j0sh,
First of all, awesome work on gensqlite! It's been super convenient to use for my projects thus far. Anyways, the package available on OPAM corresponds to the initial commit for this repo (August 12, 2015). However, there were some changes made since then that aren't officially released to OPAM. I'm not sure if they're considered incomplete (in which case I could help out keeping them up-to-date), but would you be willing to update the release on OPAM?
Thanks, nv

I see you started working on a replacement for sigils. Do you mind if I propose a syntax I like?

It is interesting or desirable to use familiar idioms, like the ${…} string interpolation. Some widespread languages (the shell, BSD Make at least) use a suffix introduced by a semicolon to add pepper and salt to a variable, such as default values or various transformations. Here are the various SQL statements found in sample.ml with a proposal for another syntax:

• insert into users(name) values(>@name) –> INSERT INTO users(name) values(${name:%s})
• select <:id, <@name, strftime('%s', <?created) from users where name = >@name –> SELECT ${id:@d}, ${name:@s}, strftime('%s', ${created:@h}) FROM users WHERE name = ${name:%s}

This also use the @ vs % convention for formatting input and output fields, as sqlexpr does.

I am currently giving gensqlite a try and it looks very promising, thanks for writing it!

This package uses internal parsetree definitions that changed in 4.03.

See the opam-builder report in http://opam.ocamlpro.com/builder/html/gensqlite/gensqlite.0.1/4e11133f38685ec7bdbb1630e8ff1c83.

Would it be possible to make a new release that is compatible with 4.03 and above?

Duplicate named params of the same type should only be bound once. (What to do if there are params with the same name but different type?)