Way to ignore built in columns

[felixmosh]

Hi,

Knex internally has some times (at migrations) usage of internal tables (such as information_schema).
Those builtin tables has special cases (CONST_CASE), and because of the casting, knex fails on migrations. (trying to access TABLE_NAME) but the object is camelCased.

There is any why to ignore those built in tables?

[Kequc]

I wasn't aware of this, could you expand on it with an example so that I understand?

[felixmosh]

When you have a migration with renameColumn, this code gets called https://github.com/knex/knex/blob/master/lib/dialects/mysql/schema/tablecompiler.js#L81 Which calls https://github.com/knex/knex/blob/master/lib/dialects/mysql/schema/tablecompiler.js#L133.

You can see in here https://github.com/knex/knex/blob/master/lib/dialects/mysql/schema/tablecompiler.js#L161 try to use TABLE_NAME column

[Kequc]

Could you give an example of what you're doing, what's happening and how you expect it to work? I'm trying to figure out what's going on and how I can fix it. I haven't ever run into the problem you are experiencing.

[felixmosh]

Create a knexfile.js with simple config.

// knexfile.js
const knexStringcase = require('knex-stringcase');

module.exports = {
  development: knexStringcase({
    client: 'mysql2',
    asyncStackTraces: true,
    connection: {
      host: process.env.MYSQL_HOST,
      database: process.env.MYSQL_DBNAME,
      user: process.env.MYSQL_USER,
      password: process.env.MYSQL_PASSWORD,
      typeCast: (field, next) => {
        if (field.type === 'TINY' && field.length === 1) {
          return field.string() === '1'; // 1 = true, 0 = false
        }
        return next();
      },
    },
    pool: {
      min: 2,
      max: 10,
    },
    migrations: {
      directory: './db/migrations',
      tableName: '__knex_migrations',
    },
    seeds: {
      directory: './db/seeds',
    },
  }),
};

Run in the terminal npx knex migrate:make --knexfile ./knexfile.js rename_columns;
This will create a new migration file.
Set it's content to:

export async function up(knex) {
  await knex.schema.alterTable('someTableName', t => {
    t.renameColumn('old_column_name', 'new_column_name');
  });
}

export async function down(knex: Knex) {
  await knex.schema.alterTable('someTableName', t => {
    t.renameColumn('new_column_name', 'old_column_name');
  });
}

Then run in the terminal npx knex migrate:latest --knexfile ./knexfile.js
This will execute in the background this internal method
which executes this, which tries to access TABLE_NAME column.

The migration will fail on alter table 'db_name.undefined' drop foreign key 'undefined'.
'undefined' is not exists in 'db_name'

[Kequc]

That's kinda interesting. The only thing my library is doing is using the post process response and wrap identifier methods provided by knex as I think they are supposed to be used. I don't have any access to knex internals nor any control over when or how knex uses these methods.

It seems like there was a similar issue posted to their repository regarding a different string mapping library.

knex/knex#3033

It says it was resolved but maybe this is a new issue?

One thing that does stand out to me a little is that you're using camel case in your table names. I think this shouldn't be a problem though, I'm curious whether or not on your database your table names are snake case.

The way knex handles using these functions is a bit of a mystery to me. On the way to the database it uses wrap identifier in order to convert strings one at a time. Then on the way back it passes the entire result as a single object.

But on the way to the database for example I don't know when it's being used. I could add an additional option for ignoring certain strings but would that be helpful?

[felixmosh]

Thank you for the quick answer!

All of my tables & columns are snake case.

Actually I don't know why it surfaced now, cause sometime ago I've used renameColumn function and it worked (I'm not sure it there was a use of stringCase).

I think that adding ignore list will allow me to ignore columns by name.
Maybe a better approach will be a PR to knex, to prevent this mapping on internal tables.

[Kequc]

Thanks I think that's the right place to ask for now. Please re-open if it comes back to me.