Metadata API Reference: Relationships

Introduction

When retrieving data from tables, it is very helpful if we can also fetch the related data alongside the columns. This is where relationships come in. They can be considered as pseudo columns for a table to access the related data.

For a simple article/author schema, the following relationships exist:

• author of an article
• articles of an author

There are two kinds of relationships:

• one-to-one or object relationships (e.g. author).
• one-to-many or array relationships (e.g. articles).

The above represents the same table relationship from different perspectives: there is a single author for every article (one-to-one), but there may be multiple articles for every author (one-to-many).

A table relationship may be one-to-one from both perspectives. For example, given tables author and author_details, if the author_details table has a primary key author_id which is a foreign key to the author table's primary key id. In this case there will be a single author for every author_details and a single details for every author

Supported from

The Metadata API is supported for versions v2.0.0 and above and replaces the older schema/Metadata API.

pg_create_object_relationship

create_object_relationship is used to create an object relationship on a table. There cannot be an existing column or relationship with the same name.

There are 3 ways in which you can create an object relationship.

1. Using foreign key constraint on a column

Create an object relationship author on article table, using the foreign_key_constraint_on the author_id column:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_create_object_relationship",
    "args": {
        "table": "article",
        "name": "author",
        "source": "default",
        "using": {
            "foreign_key_constraint_on" : ["author_id"]
        }
    }
}

Note

In the case that the key uses only a single column it is permissible to give just a string instead of a list, i.e.: "foreign_key_constraint_on" : "author_id".

2. Using foreign key constraint on a remote table

Create an object relationship details on author table, using the foreign_key_constraint_on the author_details table's author_id column:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "create_object_relationship",
    "args": {
        "table": "author",
        "name": "details",
        "using": {
            "foreign_key_constraint_on" : {
                "table": "author_details",
                "columns": ["author_id"]
            }
        }
    }
}

Deprecation

For compatibility with previous versions we also support the form of foreign_key_constraint_on with a column-field, e.g.:

{
  "foreign_key_constraint_on": {
    "table": "author_details",
    "column": "author_id"
  }
}

This form is deprecated in favor of the more general columns field.

Supported from

Relationships via remote table are supported for versions v2.0.0-alpha.3 and above.

3. Manual configuration

This is an advanced feature which is mostly used to define relationships on or to views. We cannot rely on foreign key constraints as they are not valid to or from views. So, when using manual configuration, we have to specify the remote table and how columns in this table are mapped to the columns of the remote table.

Let's say we have a view called article_detail which has three columns article_id and view_count and average_rating. We can now define an object relationship called article_detail on the article table as follows:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_create_object_relationship",
    "args": {
        "table": "article",
        "name": "article_detail",
        "source": "default",
        "using": {
            "manual_configuration" : {
                "remote_table" : "article_detail",
                "column_mapping" : {
                    "id" : "article_id"
                }
            }
        }
    }
}

Note

It is easy to make mistakes while using manual_configuration. One simple check is to ensure that foreign key constraint semantics are valid on the columns being used in column_mapping. In the previous example, if it was allowed, a foreign key constraint could have been defined on article table's id column to article_detail view's article_id column.

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
name	true	RelationshipName	Name of the new relationship
using	true	ObjRelUsing	Use one of the available ways to define an object relationship
comment	false	text	comment
source	false	SourceName	Name of the source database of the table (default: default)

pg_create_array_relationship

create_array_relationship is used to create an array relationship on a table. There cannot be an existing column or relationship with the same name.

There are 2 ways in which you can create an array relationship.

1. Using foreign key constraint on a column

Create an array relationship articles on author table, using the foreign_key_constraint_on the author_id column of the article table:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_create_array_relationship",
    "args": {
        "table": "author",
        "name": "articles",
        "source": "default",
        "using": {
            "foreign_key_constraint_on" : {
                "table" : "article",
                "columns" : ["author_id"]
            }
        }
    }
}

Deprecation

For compatibility with previous version we also support the form of foreign_key_constraint_on with a column-field, e.g.:

{
  "foreign_key_constraint_on": {
    "table": "author_details",
    "column": "author_id"
  }
}

This form is deprecated in favor of the more general columns field.

2. Manual configuration

This is an advanced feature which is mostly used to define relationships on or to views. We cannot rely on foreign key constraints as they are not valid to or from views. So, when using manual configuration, we have to specify the remote table and how columns in this table are mapped to the columns of the remote table.

Let's say we have a view called article_detail which has four columns author_id, article_id, view_count and average_rating. We can now define an array relationship called article_details on the author table as follows:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_create_array_relationship",
    "args": {
        "table": "author",
        "name": "article_details",
        "source": "default",
        "using": {
            "manual_configuration" : {
                "remote_table" : "article_detail",
                "column_mapping" : {
                    "id" : "author_id"
                }
            }
        }
    }
}

Note

It is easy to make mistakes while using manual_configuration. One simple check is to ensure that foreign key constraint semantics are valid on the columns being used in column_mapping. In the previous example, if it was allowed, a foreign key constraint could have been defined on the author table's id column to article_detail view's author_id column.

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
name	true	RelationshipName	Name of the new relationship
using	true	ArrRelUsing	Use one of the available ways to define an array relationship
comment	false	text	comment
source	false	SourceName	Name of the source database of the table (default: default)

pg_drop_relationship

pg_drop_relationship is used to drop a relationship (both object and array) on a table. If there are other objects dependent on this relationship like permissions and query templates, etc., the request will fail and report the dependencies unless cascade is set to true. If cascade is set to true, the dependent objects are also dropped.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_drop_relationship",
    "args": {
        "table": "article",
        "source": "default",
        "relationship": "article_detail"
    }
}

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
relationship	true	RelationshipName	Name of the relationship that needs to be dropped
cascade	false	Boolean	When set to true, all the dependent items on this relationship are also dropped
source	false	SourceName	Name of the source database of the table (default: default)

Note

Be careful when using cascade. First, try running the request without cascade or cascade set to false.

pg_set_relationship_comment

pg_set_relationship_comment is used to set/update the comment on a relationship. Setting the comment to null removes it.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_set_relationship_comment",
    "args": {
        "table": "article",
        "source": "default",
        "name": "article_detail",
        "comment" : "has extra information about an article like count etc."
    }
}

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
relationship	true	RelationshipName	The relationship
comment	false	Text	Comment
source	false	SourceName	Name of the source database of the table (default: default)

pg_rename_relationship

pg_rename_relationship is used to modify the name of an existing relationship.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "pg_rename_relationship",
    "args": {
        "table": "article",
        "name": "article_details",
        "source": "default",
        "new_name": "article_detail"
    }
}

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
name	true	RelationshipName	The relationship
new_name	true	RelationshipName	New relationship name
source	false	SourceName	Name of the source database of the table (default: default)

pg_suggest_relationships

pg_suggest_relationships is used to suggest relationships that can be tracked by the pg_create_*_relationship API.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
  "type": "pg_suggest_relationships",
  "version": 1,
  "args": {
    "omit_tracked": true,
    "source": "chinook",
    "tables": [["Artist"]]
  }
}

Args syntax

Key	Required	Default	Schema	Description
source	false	default	SourceName	The source to suggest relationships for
tables	false	All Tables	[ TableName ]	List of tables to consider for suggestions
omit_tracked	false	false	Boolean	Don't re-suggest tracked relationships

Note

All other source kinds are also supported by using the type {KIND}_suggest_relationships. For example, Snowflake would be: snowflake_suggest_relationships, etc.

mssql_create_object_relationship

create_object_relationship is used to create an object relationship on a table. There cannot be an existing column or relationship with the same name.

There are 3 ways in which you can create an object relationship.

1. Using foreign key constraint on a column

Create an object relationship author on article table, using the foreign_key_constraint_on the author_id column:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_create_object_relationship",
    "args": {
        "table": "article",
        "name": "author",
        "source": "default",
        "using": {
            "foreign_key_constraint_on" : ["author_id"]
        }
    }
}

Note

In the case that the key uses only a single column it is permissible to give just a string instead of a list, i.e.: "foreign_key_constraint_on" : "author_id".

2. Using foreign key constraint on a remote table

Create an object relationship details on author table, using the foreign_key_constraint_on the author_details table's author_id column:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "create_object_relationship",
    "args": {
        "table": "author",
        "name": "details",
        "using": {
            "foreign_key_constraint_on" : {
                "table": "author_details",
                "columns": ["author_id"]
            }
        }
    }
}

Deprecation

For compatibility with previous versions we also support the form of foreign_key_constraint_on with a column-field, e.g.:

{
  "foreign_key_constraint_on": {
    "table": "author_details",
    "column": "author_id"
  }
}

This form is deprecated in favor of the more general columns field.

Supported from

Relationships via remote table are supported for versions v2.0.0-alpha.3 and above.

3. Manual configuration

This is an advanced feature which is mostly used to define relationships on or to views. We cannot rely on foreign key constraints as they are not valid to or from views. So, when using manual configuration, we have to specify the remote table and how columns in this table are mapped to the columns of the remote table.

Let's say we have a view called article_detail which has three columns article_id and view_count and average_rating. We can now define an object relationship called article_detail on the article table as follows:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_create_object_relationship",
    "args": {
        "table": "article",
        "name": "article_detail",
        "source": "default",
        "using": {
            "manual_configuration" : {
                "remote_table" : "article_detail",
                "column_mapping" : {
                    "id" : "article_id"
                }
            }
        }
    }
}

Note

It is easy to make mistakes while using manual_configuration. One simple check is to ensure that foreign key constraint semantics are valid on the columns being used in column_mapping. In the previous example, if it was allowed, a foreign key constraint could have been defined on article table's id column to article_detail view's article_id column.

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
name	true	RelationshipName	Name of the new relationship
using	true	ObjRelUsing	Use one of the available ways to define an object relationship
comment	false	text	comment
source	false	SourceName	Name of the source database of the table (default: default)

mssql_create_array_relationship

create_array_relationship is used to create an array relationship on a table. There cannot be an existing column or relationship with the same name.

There are 2 ways in which you can create an array relationship.

1. Using foreign key constraint on a column

Create an array relationship articles on author table, using the foreign_key_constraint_on the author_id column of the article table:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_create_array_relationship",
    "args": {
        "table": "author",
        "name": "articles",
        "source": "default",
        "using": {
            "foreign_key_constraint_on" : {
                "table" : "article",
                "columns" : ["author_id"]
            }
        }
    }
}

Deprecation

For compatibility with previous version we also support the form of foreign_key_constraint_on with a column-field, e.g.:

{
  "foreign_key_constraint_on": {
    "table": "author_details",
    "column": "author_id"
  }
}

This form is deprecated in favor of the more general columns field.

2. Manual configuration

This is an advanced feature which is mostly used to define relationships on or to views. We cannot rely on foreign key constraints as they are not valid to or from views. So, when using manual configuration, we have to specify the remote table and how columns in this table are mapped to the columns of the remote table.

Let's say we have a view called article_detail which has four columns author_id, article_id, view_count and average_rating. We can now define an array relationship called article_details on the author table as follows:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_create_array_relationship",
    "args": {
        "table": "author",
        "name": "article_details",
        "source": "default",
        "using": {
            "manual_configuration" : {
                "remote_table" : "article_detail",
                "column_mapping" : {
                    "id" : "author_id"
                }
            }
        }
    }
}

Note

It is easy to make mistakes while using manual_configuration. One simple check is to ensure that foreign key constraint semantics are valid on the columns being used in column_mapping. In the previous example, if it was allowed, a foreign key constraint could have been defined on the author table's id column to article_detail view's author_id column.

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
name	true	RelationshipName	Name of the new relationship
using	true	ArrRelUsing	Use one of the available ways to define an array relationship
comment	false	text	comment
source	false	SourceName	Name of the source database of the table (default: default)

mssql_drop_relationship

mssql_drop_relationship is used to drop a relationship (both object and array) on a table. If there are other objects dependent on this relationship like permissions and query templates, etc., the request will fail and report the dependencies unless cascade is set to true. If cascade is set to true, the dependent objects are also dropped.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_drop_relationship",
    "args": {
        "table": "article",
        "source": "default",
        "relationship": "article_detail"
    }
}

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
relationship	true	RelationshipName	Name of the relationship that needs to be dropped
cascade	false	Boolean	When set to true, all the dependent items on this relationship are also dropped
source	false	SourceName	Name of the source database of the table (default: default)

Note

Be careful when using cascade. First, try running the request without cascade or cascade set to false.

mssql_set_relationship_comment

mssql_set_relationship_comment is used to set/update the comment on a relationship. Setting the comment to null removes it.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_set_relationship_comment",
    "args": {
        "table": "article",
        "source": "default",
        "name": "article_detail",
        "comment" : "has extra information about an article like count etc."
    }
}

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
relationship	true	RelationshipName	The relationship
comment	false	Text	Comment
source	false	SourceName	Name of the source database of the table (default: default)

mssql_rename_relationship

mssql_rename_relationship is used to modify the name of an existing relationship.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
    "type": "mssql_rename_relationship",
    "args": {
        "table": "article",
        "name": "article_details",
        "source": "default",
        "new_name": "article_detail"
    }
}

Args syntax

Key	Required	Schema	Description
table	true	TableName	Name of the table
name	true	RelationshipName	The relationship
new_name	true	RelationshipName	New relationship name
source	false	SourceName	Name of the source database of the table (default: default)

mssql_suggest_relationships

mssql_suggest_relationships is used to suggest relationships that can be tracked by the mssql_create_*_relationship API.

An example:

POST /v1/metadata HTTP/1.1
Content-Type: application/json
X-Hasura-Role: admin

{
  "type": "mssql_suggest_relationships",
  "version": 1,
  "args": {
    "omit_tracked": true,
    "source": "chinook",
    "tables": [["Artist"]]
  }
}

Args syntax

Key	Required	Default	Schema	Description
source	false	default	SourceName	The source to suggest relationships for
tables	false	All Tables	[ TableName ]	List of tables to consider for suggestions
omit_tracked	false	false	Boolean	Don't re-suggest tracked relationships