Support tweaking the catalog name

[sbernauer]

This was requested in Discord

Is there any chance to name the iceberg catalogs with underscores? Because underscores are forbidden in K8S resource names, but when using the dash sign for naming you‘ll have to use paranthesis like „iceberg-prod“.schema.table for referencing

As I was running into the exact same situation with a customer a day before I figured let's quickly address this.

Background info:

• Valid TrinoCatalog name characters are a-z and -.
• Most people want to replace the - with _, so in SQL you can use select * from my_catalog.my_schema.my_table instead of select * from "my-catalog"."my-schema"."my-table"

I'm proposing two small new features:

1. Add override mechanism for maximum flexibility.
2. As 1. is error-prone (we need to raise an error in case the same catalog name is used in different TrinoCatalogs) and requires duplication in the TrinoCatalog, we offer a helper to replace - with _. As it it would be a breaking change, we sadly need to make it opt-in.

Both options are super easy to implement. They give easy replacement as well as maximum flexibility to the user.

I quickly spiked this and ended up with the following CRD change

# Top level properties on TrinoCatalog
catalogNameOverride:
  description: |-
    Override the name of the catalog as it shows up in Trino.

    By default, the Trino catalog is named the same as the `.metadata.name` of the
    TrinoCatalog resource. It is your responsibility to make sure that no catalog names
    clash, the operator will raise an error in that case.

    TIP: If you only want to replace `-` with `_` use `catalogNameReplaceHyphens` instead.
  nullable: true
  type: string
catalogNameReplaceHyphens:
  default: false
  description: |-
    Wether hyphens (`-`) in the name of the catalog should be replaced by underscores (`_`).

    This is recommended because Kubernetes only allows `a-z` and `-`, while Trino
    requires quoting for catalogs containing `-` characters, but not for `_`. This mechanism
    allows you to use valid Kubernetes names, but keeps the convenience of `_` in catalog
    names.

    In case you need complete flexibility over the catalog name, you can use
    `catalogNameOverride`.
  type: boolean

[Jimvin]

• adjustCatalogName would have unexpected side effects for catalogs where the user intends it to have a hyphen character in the catalog name. Opt-in would be preferable.
• Would catalogName be a better property name than catalogNameOverride? Make it optional and set catalogName = metadata.name if it is not defined.
• The CRD should enforce that only legal names for Trino catalogs be used and maybe emit a warning if it contains problematic characters.

[sbernauer]

• I'm pretty sure most of the users that have hypens would prefer them to show up as underscores (as quoting is really annoying). If I'd need to guess, I'd say 90% maybe, do you have a different estimate? That's why I tried to model the default to fit the most cases possible.
• No strong opinion, I would like other opinions as well :) Overrides matches the "ignore what the operator calculated, just use this" terminology and makes it clear that you can shoot yourself in the foot. On the other hand catalogName is more conceive.
• That should already be the case and should not change when inferring it from the TrinoCatalog name: https://kubernetes.io/docs/concepts/overview/working-with-objects/names/#dns-subdomain-names, e.g. contain only lowercase alphanumeric characters, '-' or '.'. Note: Technically it's allowed to contain a dot, but things go boom for unrelated reasons.

[lfrancke]

• I'm also against adding adjustCatalogName at all
  • It can also have unexpected side-effects should we change the behavior later.
  • I found it confusing
• No strong opinion either