GraphQL API

This guide explains the GraphQL Query API that is used for the Graph Protocol.

Queries

In your subgraph schema you define types called Entities. For each Entity type, an entity and entities field will be generated on the top-level Query type. Note that query does not need to be included at the top of the graphql query when using The Graph.

Examples

Query for a single Token entity defined in your schema:

{
  token(id: "1") {
    id
    owner
  }
}

Note: When querying for a single entity, the id field is required and it must be a string.

Query all Token entities:

{
  tokens {
    id
    owner
  }
}

Sorting

When querying a collection, the orderBy parameter may be used to sort by a specific attribute. Additionally, the orderDirection can be used to specify the sort direction, asc for ascending or desc for descending.

Example

{
  tokens(orderBy: price, orderDirection: asc) {
    id
    owner
  }
}

Pagination

When querying a collection, the first parameter can be used to paginate from the beginning of the collection. It is worth noting that the default sort order is by ID in ascending alphanumeric order, not by creation time.

Further, the skip parameter can be used to skip entities and paginate. e.g. first:100 shows the first 100 entities and first:100, skip:100 shows the next 100 entities.

Queries should avoid using very large skip values since they generally perform poorly. For retrieving a large number of items, it is much better to page through entities based on an attribute as shown in the last example.

Example using first

Query the first 10 tokens:

{
  tokens(first: 10) {
    id
    owner
  }
}

To query for groups of entities in the middle of a collection, the skip parameter may be used in conjunction with the first parameter to skip a specified number of entities starting at the beginning of the collection.

Example using first and skip

Query 10 Token entities, offset by 10 places from the beginning of the collection:

{
  tokens(first: 10, skip: 10) {
    id
    owner
  }
}

Example using first and id_ge

If a client needs to retrieve a large number of entities, it is much more performant to base queries on an attribute and filter by that attribute. For example, a client would retrieve a large number of tokens using this query:

query manyTokens($lastID: String) {
  tokens(first: 1000, where: { id_gt: $lastID }) {
    id
    owner
  }
}

The first time, it would send the query with lastID = "", and for subsequent requests would set lastID to the id attribute of the last entity in the previous request. This approach will perform significantly better than using increasing skip values.

Filtering

You can use the where parameter in your queries to filter for different properties. You can filter on mulltiple values within the where parameter.

Example using where

Query challenges with failed outcome:

{
  challenges(where: { outcome: "failed" }) {
    challenger
    outcome
    application {
      id
    }
  }
}

You can use suffixes like _gt, _lte for value comparison:

Example for range filtering

{
  applications(where: { deposit_gt: "10000000000" }) {
    id
    whitelisted
    deposit
  }
}

Example for block filtering

You can also filter entities by the _change_block(number_gte: Int) - this filters entities which were updated in or after the specified block.

This can be useful if you are looking to fetch only entities which have changed, for example since the last time you polled. Or alternatively it can be useful to investigate or debug how entities are changing in your subgraph (if combined with a block filter, you can isolate only entities that changed in a specific block).

{
  applications(where: { _change_block: { number_gte: 100 } }) {
    id
    whitelisted
    deposit
  }
}

Example for nested entity filtering

Filtering on the basis of nested entities is possible in the fields with the _ suffix.

This can be useful if you are looking to fetch only entities whose child-level entities meet the provided conditions.

{
  challenges(where: { application_: { id: 1 } }) {
    challenger
    outcome
    application {
      id
    }
  }
}

All Filters

Full list of parameter suffixes:

_
_not
_gt
_lt
_gte
_lte
_in
_not_in
_contains
_contains_nocase
_not_contains
_not_contains_nocase
_starts_with
_starts_with_nocase
_ends_with
_ends_with_nocase
_not_starts_with
_not_starts_with_nocase
_not_ends_with
_not_ends_with_nocase

Please note that some suffixes are only supported for specific types. For example, Boolean only supports _not, _in, and _not_in, but _ is available only for object and interface types.

In addition, the following global filters are available as part of where argument:

_change_block(number_gte: Int)

Time-travel queries

You can query the state of your entities not just for the latest block, which is the default, but also for an arbitrary block in the past. The block at which a query should happen can be specified either by its block number or its block hash by including a block argument in the toplevel fields of queries.

The result of such a query will not change over time, i.e., querying at a certain past block will return the same result no matter when it is executed, with the exception that if you query at a block very close to the head of the Ethereum chain, the result might change if that block turns out to not be on the main chain and the chain gets reorganized. Once a block can be considered final, the result of the query will not change.

Note that the current implementation is still subject to certain limitations that might violate these gurantees. The implementation can not always tell that a given block hash is not on the main chain at all, or that the result of a query by block hash for a block that can not be considered final yet might be influenced by a block reorganization running concurrently with the query. They do not affect the results of queries by block hash when the block is final and known to be on the main chain. This issue explains what these limitations are in detail.

Example

{
  challenges(block: { number: 8000000 }) {
    challenger
    outcome
    application {
      id
    }
  }
}

This query will return Challenge entities, and their associated Application entities, as they existed directly after processing block number 8,000,000.

Example

{
  challenges(block: { hash: "0x5a0b54d5dc17e0aadc383d2db43b0a0d3e029c4c" }) {
    challenger
    outcome
    application {
      id
    }
  }
}

This query will return Challenge entities, and their associated Application entities, as they existed directly after processing the block with the given hash.

Fulltext Search Queries

Fulltext search query fields provide an expressive text search API that can be added to the subgraph schema and customized.

Fulltext search queries have one required field, text, for supplying search terms. Several special fulltext operators are available to be used in this text search field.

Fulltext search operators:

Symbol	Operator	Description
&	And	For combining multiple search terms into a filter for entities that include all of the provided terms
|	Or	Queries with multiple search terms separated by the or operator will return all entities with a match from any of the provided terms
<->	Follow by	Specify the distance between two words.
:*	Prefix	Use the prefix search term to find words whose prefix match (2 characters required.)

Examples

Using the or operator, this query will filter to blog entities with variations of either "anarchism" or "crumpet" in their fulltext fields.

{
  blogSearch(text: "anarchism | crumpets") {
    id
    title
    body
    author
  }
}

The follow by operator specifies a words a specific distance apart in the fulltext documents. The following query will return all blogs with variations of "decentralize" followed by "philosophy"

{
  blogSearch(text: "decentralized <-> philosophy") {
    id
    title
    body
    author
  }
}

Combine fulltext operators to make more complex filters. With a pretext search operator combined with a follow by this example query will match all blog entities with words that start with "lou" followed by "music".

{
  blogSearch(text: "lou:* <-> music") {
    id
    title
    body
    author
  }
}

Validation

Graph Node implements specification-based validation of the GraphQL queries it receives using graphql-tools-rs, which is based on the graphql-js reference implementation. Queries which fail a validation rule do so with a standard error - visit the GraphQL spec to learn more.

Schema

The schema of your data source--that is, the entity types, values, and relationships that are available to query--are defined through the GraphQL Interface Definition Langauge (IDL).

GraphQL schemas generally define root types for queries, subscriptions and mutations. The Graph only supports queries. The root Query type for your subgraph is automatically generated from the GraphQL schema that's included in your subgraph manifest.

Note: Our API does not expose mutations because developers are expected to issue transactions directly against the underlying blockchain from their applications.

Entities

All GraphQL types with @entity directives in your schema will be treated as entities and must have an ID field.

Note: Currently, all types in your schema must have an @entity directive. In the future, we will treat types without an @entity directive as value objects, but this is not yet supported.

Subgraph Metadata

All subgraphs have an auto-generated _Meta_ object, which provides access to subgraph metadata. This can be queried as follows:

{
  _meta(block: { number: 123987 }) {
    block {
      number
      hash
      timestamp
    }
    deployment
    hasIndexingErrors
  }
}

If a block is provided, the metadata is as of that block, if not the latest indexed block is used. If provided, the block must be after the subgraph's start block, and less than or equal to the most recently indexed block.

deployment is a unique ID, corresponding to the IPFS CID of the subgraph.yaml file.

block provides information about the latest block (taking into account any block constraints passed to _meta):

• hash: the hash of the block
• number: the block number
• timestamp: the timestamp of the block, if available (this is currently only available for subgraphs indexing EVM networks)

hasIndexingErrors is a boolean identifying whether the subgraph encountered indexing errors at some past block