Introspect a PostgreSQL database to generate Go marshalling boilerplate

go get -u github.com/wttw/mro

mro --bootstrap pgx will create four files in the current directory, suitable for use with pgx.

mro.cfg is a HCL format configuration file. It's hopefully self-documenting. If nothing else you'll need to edit the ConnectionString setting to point at the database containing the schema you want to work from.

pgx.go specifies the interface that mro generated code will use to access the database. It's implemented by pgx.Conn, pgx.ConnPool and pgx.Tx.

table.pgx.tpl and enum.pgx.tpl are Go format templates used to generate code.

mro or mro -package <packagename> will generate marshaling and unmarshaling code for the database schema. For each table it will generate a struct that represents a row of the table, with a name based on the name of the table converted into PascalCase: a table called "email_source" will map on to a struct called "EmailSource". That struct has an Insert() method and, if there's a single-column primary key, Update(), Upsert() and Delete() methods.

Functions to retrieve data from each table are also created. AllEmailSource() will return the entire table, and functions named like EmailSourceByID() will be created for each primary key or unique index on the table.

Additional SQL queries can be added to the Queries section of the configuration file. These must retrieve columns from a single table, and will generate functions to retrieve those as slices of that table's struct.

It will also generate mro.json containing all the information retrieved from the database, as is passed in to the templates to generate code.

Any database other than PostgreSQL.

Marshaling of arbitrary SQL queries. This could be added fairly simply, but I've not needed it so far.

Any use of, for example, foreign keys to record structure at a higher level than a table.

Anything ORM-ish, beyond basic marshaling code. If you're looking for programatic generation of SQL queries or automatic loading of related rows this isn't the best place to start.

Table or column names that use unicode characters outside the basic multilingual plane aren't handled correctly.

This is mostly untested code. I'm using it in a large production-grade project, so I'll be dealing with bugs (and maybe adding regression test) as I come across them.

It's also a fairly quick hack, so while the code quality isn't terrible it's definitely not great. Global variables, everything lives in main ...

gnorm

xo

sqlboiler - much more ORMy