sanitize/validate service name input

[n1ru4l]

We should validate and sanitize the service name for schema publishes and schema checks. Right now the service string can be anything, including whitespace and special characters.

We should limit it more to have:

• maximum length
• disallow spaces
• disallow certain special characters

We need to consider already existing services published to our database and ensure that introducing the limits does not break any existing users.

Characters used within our pg database today:

 "-./0123456789_abcdefghijklmnopqrstuvwxyz{}“”

how we extracted the list

select string_agg(c,'')
from (
  select distinct regexp_split_to_table(lower(service_name),'') as c
  from schema_log
) t

Some of those look like unintended user mistakes when publishing the schema:

• { and } seem like wrongly interpolated environment variables on CI systems (projects using this are not active; or all services deleted)
• ” and ” also seem like wrongly interpolated environment variables on CI systems (project using this are not active; or all services deleted)

Proposed allowed characters:

• alphanumeric
• dots
• underscore
• slash
• dash

[jdolle]

For now I think we should only validate when creating a new service. We can then work to migrate existing services if we need to. But I don't want to break anyone's services if for some reason they are using one of these characters.