Hacker News Reader

Comment by sgarland | Hacker News Reader

YAML is bad, first and foremost, because the language is hideously awful. Anything that can interpret Norway’s country code as a bool, or certain numbers as sexagesimal, has not been well-thought out.

As to what else, I continually fail to see what’s wrong with SQL. It’s incredibly easy to read, in plain English, and understand what is being done. ALTER TABLE foo ADD COLUMN bar INTEGER NOT NULL DEFAULT 0… hmm, wonder what that’s doing?

For people who say you get too many .sql files doing it this way: a. Get your data models right the first time b. Build a tool to parse and combine statements, if you must.

View on HN

I agree with your point about SQL's readability, however, I think there's value in having structured metadata around migrations, like dependencies between migrations, rollback instructions, and environment specific variations, beyond just the raw SQL.

This could be done with SQL comments, but having it in a structured format makes it more reliable to parse and validate programmatically. I do see why YAML's quirks could become a problem in a tool that's meant to help you make sure your database is in order, we didn't run into issues like country codes or numbers being interpreted as sexagesimal (yet).

Perhaps a middle ground would be to keep the actual migrations in .sql files for readability, while using a separate metadata file (in JSON or TOML) for the orchestration details. What do you think?

Re: dependencies, SQL is declarative. While there might be benefit in keeping the last modification as its own entity, I think just dumping the output of SHOW CREATE TABLE or the equivalent is the best solution. It clearly shows the current state of the table. This of course requires that the migration has already occurred, or has been faked for display purposes.

I guess my stance is that if you need some kind of config file to create the migration, then fine, but as a DBRE, it’s much easier for me to reason about the current state by seeing the declarative output than by mentally summing migration files. As a bonus, this lets you see quite easily if you’ve done something silly like creating an additional index on a column that has a UNIQUE constraint (though you’d still have to know why that’s unnecessary, I suppose).

Yep that's the reason we generate a schema.sql at the end so that you know the current state of the database after the migrations have been applied (you don't need to mentally sum those migrations). Coupled with git, you get to see the diff of the schema.sql to see what changed which is super useful, like the same additional index example you mentioned.

Reply on news.ycombinator.com