Avram::Migrator::Runner.new.run_pending_migrations not work anymore?

[zw963]

Hi, i am back!

What i want is, to check/run migration when the app is running, instead of use lucky binary.

Then, i add following code into the last line of a new 1.1.0 project's config/database.cr

if !LuckyEnv.task?
  Avram::Migrator::Runner.new.run_pending_migrations
end

I found those code never work, because Avram::Migrator::Runner.migrations always return a empty list, l always get empty migrations.

sorted_migrations # => self.class.migrations # => []
[]
self.class.migrations # => []
Did not migrate anything because there are no pending migrations.

Maybe because I haven't used the Lucky/avram for a while now, but, i remember when i start o use avram with kemal, i can achieve this.

I guess, maybe i missing some config which search migrations file in db/migrations folder? there is 00000000000001_create_users.cr there, i still not create user table yet.

Thanks!

[zw963]

Okay, add a require "../db/migrations/**" into config/database.cr do the trick.

So, is there possible to create a precompile binary, like bin/lucky.db.migrate?

Thanks.

[robacarp]

A lucky app provides tasks.cr, which is what eventually gets executed when you call lucky db.migrate from your terminal. If you add it to your shard.yml as a target:

targets:
  tasks:
    main: tasks.cr

Then when you deploy with a shards build, you'll get an executable at bin/tasks which you can use to run migrations with bin/tasks db.migrate

[zw963]

Okay, add a require "../db/migrations/**" into config/database.cr do the trick.

Hi, i know the reason i have to add require "../db/migration/**" into config/database.cr for solve my issue.

because the require order in the src/app.cr cause this issue.

require "./shards"

# ...
require "../config/**"

# ...
require "../db/migrations/**"
require "./app_server"

Is there special reason we can't write the "require ../db/migrations/* *" earlier? just before the require "../config"?

Or, just like how I have to manually modify every time, add it into require "../config/database.cr"?

+ require "../db/migrations/**"

# ...

+ if !LuckyEnv.task?
+   Avram::Migrator::Runner.new.run_pending_migrations
+ end

[jwoertink]

I think as long as your migrations are required after your models, it should be fine.

# the migration would need to know that Post and User exists
create table_for(Post) do
  # ...
  add_belongs_to user : User
end

But since those are used in a Macro, and not actually resolved, then maybe you could even move the require to after require "avram"?

[zw963]

But since those are used in a Macro, and not actually resolved, then maybe you could even move the require to after require "avram"?

I think so too.

The issue is, where to add Avram::Migrator::Runner.new.run_pending_migrations, okay, please let me explain why i need check pending migrations in my deployment process. (anyway, I think the process introduced in official doc is a bit complicated)

1. yard proc for package assets into public folder
2. use baked_file_system_mounter to add assets files in public into binary when build, and auto mount it on production.
3. build static locally use zig cc. there is a toturial for how to do this, then i have a big static binary, which is ready for copy to production.
4. copy file into production, and run it, done.
5. it will mount assets file into specified folder (public), and check the pending migration, run migration if necessary. no need crystal or nodejs installed on server, deploy is easy.

The only issue is, where i add Avram::Migrator::Runner.new.run_pending_migrations into? the prefer place is config/database.cr, but add to that end of that file not work, so, i have to add require "../db/migrations/**" at the beginning of config/database.cr too solve this issue.

In fact, i tried, remove require "../db/migrations/**" fromsrc/app.cr, and add those back into config/database.cr`, it work, and this refactor seem like make sense.

Thanks

[jwoertink]

All of my apps use Avram::Migrator::Runner.new.run_pending_migrations on deploy too. I put my require migrations in app.cr and the method call goes in start_server.cr

require "./app"

Habitat.raise_if_missing_settings!

if LuckyEnv.development?
  Avram::Migrator::Runner.new.ensure_migrated!
  Avram::SchemaEnforcer.ensure_correct_column_mappings!
else
  Avram::Migrator::Runner.new.run_pending_migrations
end

app_server = AppServer.new

Signal::INT.trap do
  app_server.close
end

app_server.listen

[zw963]

I put my require migrations in app.cr

Yes, i thought this is the key point. migration is related to database, add it into config/database.cr is better?

Or, if anyone use avaram without lucky, where the run_pending_migrations should add to? i guess still should be config/database.cr

Thanks