Reserved Keywords / Table Names
Some table, column, function and argument names can cause issues with PostGraphile due to conflicts with auto-generated type and field names. In these cases you can generally change the GraphQL type and field names for your PostgreSQL schema to avoid conflicts using smart comments.
Note that in many cases the singular or the plural of the words below may also cause issues, so assume the list below includes singulars and plurals too. The same goes for snake_case vs camelCase since PostGraphile automatically converts between these for you.
- [Table]
query
,mutation
,subscription
- conflict with the root-levelQuery
,Mutation
andSubscription
GraphQL types. - [Table, Function]
node
- conflicts with the GraphQL rootnode(...)
interface. - [Column, Argument]
order_by
/orderBy
- Used by PostGraphile for ordering, may cause issues when declared as a function argument or table column. - [Column]
primary_key
- we automatically addorderBy: PRIMARY_KEY_ASC
to ordering, if you have a column calledprimary_key
then this value will be generated twice and cause issues.
There are other potential conflicts too, if you discover more conflicts then please consider using the "Edit this page" link above to suggest some changes!
Non-unique table names
If there exists multiple schemas with the same table name inside of them, it can
cause issues with the nodeId: ID
(or id: ID
) field, by making IDs which are
not globally unique. Calls utilizing those IDs may return unexpected results. In
order to avoid this, we must disable the PgNodeAliasPostGraphile
plugin, and
therefore use the GraphQL type name rather than the PostgreSQL table name in the
identifiers. On the CLI you'd do this like:
postgraphile --skip-plugins graphile-build-pg:PgNodeAliasPostGraphile
.
To support non-unique table names you should implement an inflection override that will change how the GraphQL type names are generated from the raw table name.