OpenAPI
If data is returned in JSON Lines format (*.jsonl) then the schema applies to the JSON object obtained by combining
the sub-objects accessed via the key and value properties, respectively, of JSON items.
Assume the schema reads as follows:
{
"type": "object",
"properties": {
"pkey": {
"type": "integer",
"format": "int64"
},
"prop1": {
"type": "string"
},
"prop2": {
"type": "integer"
},
"additionalProperties": false,
"required": [
"pkey",
"prop1"
]
}
}
Suppose we have the following JSON output:
{ "meta": { "action": "U", ... }, "key": { "pkey": 1 }, "value": { "prop1": "value1", "prop2": 42 } }
{ "meta": { "action": "U", ... }, "key": { "pkey": 2 }, "value": { "prop1": "value2", "prop2": null } }
{ "meta": { "action": "D", ... }, "key": { "pkey": 3 } }
In the example directly above, the first and second items (update records) would both validate against the pre-defined schema.
The validator would check the following synthesized JSON objects:
{ "pkey": 1, "prop1": "value1", "prop2": 42 }
{ "pkey": 2, "prop1": "value2", "prop2": null }
The third item (a delete record) does not have to validate because it indicates that the client is to remove the item.
If data is returned in Comma-Separated Values format (*.csv) then the schema type constraints apply to CSV key and value
columns, respectively, but not CSV meta columns. For example, assume we have the following CSV output:
meta.action,key.pkey,value.prop1,value.prop2
U,1,"value1",42
U,2,"value2",
D,3,,
Then the schema would read the same as in the JSON example above.
Nested JSON objects are flattened to simple fields, with composite names constructed using the dot notation (parent.child).
If no scope is specified, then the endpoint uses the default scope of the authenticated user. Returns an error if the user has access to several scopes and the scope is not explicitly specified.
Identifies the domain or product that the request pertains to, e.g. canvas.
Canonical name of the table whose schema to return.
Identifies the scope to access, e.g. a root account UUID for Canvas, or a district ID for Mastery.
GET /dap/query/{namespace}/table/{table}/schema HTTP/1.1
Host: api-gateway.instructure.com
Authorization: Bearer JWT
Accept: */*
{
"schema": {
"type": "object",
"properties": {
"id": {
"type": "integer",
"format": "int64",
"title": "The ID of this version object. Primary key."
},
"user_id": {
"type": "integer",
"format": "int64",
"title": "The ID of the submitter. Foreign key to `users.id`."
},
"context_id": {
"type": "integer",
"format": "int64",
"title": "The ID of the course this submission belongs to. Foreign key to `courses.id`."
},
"context_type": {
"type": "string",
"enum": [
"Course"
],
"title": "The type of the context object (typically `Course`)."
}
},
"additionalProperties": false,
"required": [
"id",
"context_type"
],
"title": "This table stores information describing previous versions of individual submission objects."
},
"version": 1
}In contrast to objects, which have a longer lifetime, pre-signed URLs are valid for a shorter duration, typically 15 minutes.
File paths returned by this endpoint do not adhere to any specification. While they may contain auxiliary information such as job ID or part counter, these are only informative. Downstream systems should not depend on any specific patterns of file names, or make any assumptions how much data each file contains.
The lifetime of the object depends on the operation that created it but typically lasts for 24 hours. Object identifiers can be traded for pre-signed URLs via an authenticated endpoint operation while the object exists.
Uniquely identifies the object.
POST /dap/object/url HTTP/1.1
Host: api-gateway.instructure.com
Authorization: Bearer JWT
Content-Type: application/json
Accept: */*
Content-Length: 33
[
{
"id": "uuid-1"
},
{
"id": "uuid-2"
}
]{
"urls": {
"ANY_ADDITIONAL_PROPERTY": {
"url": "https://example.com"
}
}
}This is an asynchronous operation. Calling this endpoint will start a new job and return immediately with status information. However, the operation will continue running on the server. The caller can poll the status of the job to find out when it is ready.
If a job with the same query parameters already exists, its details are returned rather than starting a new job.
If no scope is specified, then the endpoint uses the default scope of the authenticated user. Returns an error if the user has access to several scopes and the scope is not explicitly specified.
For incremental queries, the output uses a special metadata field called action to identify whether a record is upserted
(inserted or updated) or (hard) deleted (U corresponds to upsert, and D to delete):
{ "meta": { "action": "U", ... }, "key": { "pkey": 1 }, "value": { "prop1": "value1", "prop2": 42 } }
{ "meta": { "action": "U", ... }, "key": { "pkey": 2 }, "value": { "prop1": "value2", "prop2": null } }
{ "meta": { "action": "D", ... }, "key": { "pkey": 3 } }
Upserted records have the primary key fields present in the sub-object key and all other data fields in the sub-object value.
Deleted records only have the primary key fields in the key property, and lack the value property.
Hard deletes are infrequent. They only take place when a record is irreversibly deleted from the source database, e.g. to comply with privacy or legal requirements.
In most cases, records are soft-deleted, i.e. they are updated in such a way as to be understood deleted or inactive though
the record is retained in the database, e.g. by setting a workflow_state column to the value inactive or deleted. In this
context, soft deletes are equivalent to an update, and are denoted with a U, and all field values are included in the output.
In the rare event that inserting a record is quickly followed by a hard delete in the source database between two successive
incremental queries, a record might appear with a new (so far unseen) key, no value and an action of D.
For snapshot queries, deleted records are not included in the output:
{ "meta": { ... }, "key": { "pkey": 1 }, "value": { "prop1": "value1", "prop2": 42 } }
{ "meta": { ... }, "key": { "pkey": 2 }, "value": { "prop1": "value2", "prop2": null } }
This is a rate-limited endpoint. If excessive data volume is requested repeatedly using this endpoint (e.g. a full snapshot every hour), future requests may be denied. We encourage making use of incremental queries, which substantially reduce the amount of data returned.
Difference between snapshot and incremental Queries:
Snapshot queries and incremental queries serve different purposes in data retrieval and management within the DAP environment.
- Snapshot Queries: These queries generate a complete and comprehensive snapshot of the entire dataset at a given point in time. However, note that for the canvas_logs dataset, particularly the web_logs table, there is a 30-day data retention policy, so the snapshot will not cover the entire dataset but only the last 30 days of data. Snapshot queries are ideal for creating an initial full copy of the dataset or for occasional full updates. It is not recommended to request snapshots regularly, as it is resource-intensive on the API side and expensive to process on the client side. This approach ensures that you have a full, standalone version of the data, which can be useful for comprehensive analyses, audits, and backups.
- Incremental Queries: In contrast, incremental queries retrieve only the changes or updates that have occurred since the last query. This method is more efficient and resource-effective as it minimizes data transfer by only fetching new or modified records. Incremental queries are ideal for keeping your dataset up-to-date with minimal overhead, enabling near-real-time data updates and reducing the need for frequent full dataset downloads.
It is recommended taking a snapshot exactly once (as an initialization step) and then using incremental queries thereafter. By utilizing snapshot queries for initial data load and incremental queries for subsequent updates, users can maintain up-to-date datasets efficiently and effectively.
Identifies the domain or product that the request pertains to, e.g. canvas.
Canonical name of the table whose data to return.
Identifies the scope to access, e.g. a root account UUID for Canvas, or a district ID for Mastery.
POST /dap/query/{namespace}/table/{table}/data HTTP/1.1
Host: api-gateway.instructure.com
Authorization: Bearer JWT
Content-Type: application/json
Accept: */*
Content-Length: 80
{
"format": "jsonl",
"since": "2022-07-14T10:30:00Z",
"until": "2022-07-14T14:30:00Z"
}{
"id": "text",
"status": "waiting",
"expires_at": "2025-07-02T00:15:50.907Z",
"objects": [
{
"id": "text"
}
],
"schema_version": 1,
"at": "2025-07-02T00:15:50.907Z"
}If no scope is specified, then the endpoint uses the default scope of the authenticated user. Returns an error if the user has access to several scopes and the scope is not explicitly specified.
Identifies the domain or product that the request pertains to, e.g. canvas.
Identifies the scope to access, e.g. a root account UUID for Canvas, or a district ID for Mastery.
GET /dap/query/{namespace}/table HTTP/1.1
Host: api-gateway.instructure.com
Authorization: Bearer JWT
Accept: */*
{
"tables": [
"accounts",
"submissions",
"users"
]
}If data is returned in JSON Lines format (*.jsonl) then the schema applies to the JSON object obtained by combining
the sub-objects accessed via the key and value properties, respectively, of JSON items.
Assume the schema reads as follows:
{
"type": "object",
"properties": {
"pkey": {
"type": "integer",
"format": "int64"
},
"prop1": {
"type": "string"
},
"prop2": {
"type": "integer"
},
"additionalProperties": false,
"required": [
"pkey",
"prop1"
]
}
}
Suppose we have the following JSON output:
{ "meta": { "action": "U", ... }, "key": { "pkey": 1 }, "value": { "prop1": "value1", "prop2": 42 } }
{ "meta": { "action": "U", ... }, "key": { "pkey": 2 }, "value": { "prop1": "value2", "prop2": null } }
{ "meta": { "action": "D", ... }, "key": { "pkey": 3 } }
In the example directly above, the first and second items (update records) would both validate against the pre-defined schema.
The validator would check the following synthesized JSON objects:
{ "pkey": 1, "prop1": "value1", "prop2": 42 }
{ "pkey": 2, "prop1": "value2", "prop2": null }
The third item (a delete record) does not have to validate because it indicates that the client is to remove the item.
If data is returned in Comma-Separated Values format (*.csv) then the schema type constraints apply to CSV key and value
columns, respectively, but not CSV meta columns. For example, assume we have the following CSV output:
meta.action,key.pkey,value.prop1,value.prop2
U,1,"value1",42
U,2,"value2",
D,3,,
Then the schema would read the same as in the JSON example above.
Nested JSON objects are flattened to simple fields, with composite names constructed using the dot notation (parent.child).
If no scope is specified, then the endpoint uses the default scope of the authenticated user. Returns an error if the user has access to several scopes and the scope is not explicitly specified.
Identifies the domain or product that the request pertains to, e.g. canvas.
Canonical name of the table whose schema to return.
Identifies the scope to access, e.g. a root account UUID for Canvas, or a district ID for Mastery.
GET /dap/query/{namespace}/table/{table}/schema HTTP/1.1
Host: api-gateway.instructure.com
Authorization: Bearer JWT
Accept: */*
{
"schema": {
"type": "object",
"properties": {
"id": {
"type": "integer",
"format": "int64",
"title": "The ID of this version object. Primary key."
},
"user_id": {
"type": "integer",
"format": "int64",
"title": "The ID of the submitter. Foreign key to `users.id`."
},
"context_id": {
"type": "integer",
"format": "int64",
"title": "The ID of the course this submission belongs to. Foreign key to `courses.id`."
},
"context_type": {
"type": "string",
"enum": [
"Course"
],
"title": "The type of the context object (typically `Course`)."
}
},
"additionalProperties": false,
"required": [
"id",
"context_type"
],
"title": "This table stores information describing previous versions of individual submission objects."
},
"version": 1
}Last updated