Why is my postgres constraint name suffix different from the standard?

In this SO answer:
PostgreSQL: default constraint names

It would appear that my foreign key (not named at time of table creation) should have defaulted to the suffix _fkey…
But it seems to have instead defaulted to the suffix _foreign instead. Is this normal?

I am worried about creating a migration where I assume the constraint name and it works in my environment and then breaks everywhere else…

Is there way to ensure I reference the constraint correctly in any environment?

Updated with info:

• posgres version: 16.4

• client that created the key: knex

• The complete FK command used: table.foreign('userId').references('users.id')

• psql with \d:

  I’m sorry but I am not allowed to post code from our codebase, so here it is with redactions:

Indexes:
    "***_pkey" PRIMARY KEY, btree (id)
    "***" UNIQUE CONSTRAINT, btree ("***Id", "***Id")
Foreign-key constraints:
    "***_***id_foreign" FOREIGN KEY ("***Id") REFERENCES ***(id)
    "***_userid_foreign" FOREIGN KEY ("userId") REFERENCES users(id)    <=== the column of interest

>Solution :

https://knexjs.org/guide/schema-builder.html#foreign shows the syntax:

table.foreign(columns, [foreignKeyName])
[.onDelete(statement).onUpdate(statement).withKeyName(foreignKeyName).deferrable(type)]

This appears to show an optional method withKeyName() you can use to specify the constraint name.

I’m not a user of knex, so I’m not going to test this. Give it a try!

Edit: If you don’t specify the constraint name, knex chooses one, and it may not be the same as PostgreSQL’s own default. That is, the knex framework does specify a constraint name, whether it’s one you specify or not.

You noticed the documentation states this:

A default key name using the columns is used unless foreignKeyName is specified.