Contest API

This is the development draft for the Contest API. See also the version that will be used at WF 2020.

Introduction

This page describes an API for accessing information provided by a Contest Control System or Contest Data Server. Such an API can be used by a multitude of clients:

• an external scoreboard
• a scoreboard resolver application
• contest analysis software, such as the ICAT toolset
• another "shadow" CCS, providing forwarding of submissions and all relevant information
• internally, to interface between the CCS server and judging instances

This API is meant to be useful, not only at the ICPC World Finals, but more generally in any ICPC-style contest setup. It is meant to incorporate and supersede a number of deprecated or obsolete specifications amongst which the JSON Scoreboard, the REST interface for source code fetching and the Contest start interface.

This REST interface is specified in conjunction with a new NDJSON event feed, which provides all changes to this interface as CRUD-style events and is meant to supersede the old XML Event Feed.

General design principles

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119.

The interface is implemented as a HTTP REST interface that outputs information in JSON format (RFC 7159). This REST interface should be provided over HTTPS to guard against eavesdropping on sensitive contest data and authentication credentials (see roles below).

Endpoint URLs

The specific base URL of this API will be dependent on the server (e.g. main CCS or CDS) providing the service; in the specification we only indicate the relative paths of API endpoints with respect to a baseurl. In all the examples below the baseurl is https://example.com/api.

We follow standard REST practices so that a whole collection can be requested, e.g. at the URL path

GET https://example.com/api/contests/wf14/teams

while an element with specific ID is requested as

GET https://example.com/api/contests/wf14/teams/10

A collection is always returned as a JSON list of objects. Every object in the list represents a single element (and always includes the ID). When requesting a single element the exact same object is returned. E.g. the URL path

GET baseurl/collection

returns

[ { "id":<id1>, <element specific data for id1> },
  { "id":<id2>, <element specific data for id2> },
     ...
]

while the URL path

GET baseurl/<collection>/<id1>

returns

{ "id":<id1>, <element specific data for id1> }

HTTP headers

A server should allow cross-origin requests by setting the Access-Control-Allow-Origin HTTP header:

Access-Control-Allow-Origin: *

A server should specify how clients should cache file downloads by setting the Cache-Control or Expires HTTP headers:

Cache-Control: public, max-age=3600, s-maxage=18000

Expires: Wed, 18 Jul 2018 07:28:00 GMT

HTTP methods

The current version of this specification only requires support for the GET method, unless explicitly specified otherwise in an endpoint below (see e.g. PATCH start_time). However, for future compatibility below are already listed other methods with their expected behavior, if implemented.

• GET Read data. This method is idempotent and does not modify any data. It can be used to request a whole collection or a specific element.

• POST Create a new element. This can only be called on a collection endpoint, and the id attribute may not be specified as it is up to the server to assign one. If successful the response will contain a Location header pointing to the newly created element.

• PUT Creates or replaces a specific element. This method is idempotent, can only be called on a specific element, and replaces its contents with the data provided. The payload data must be complete, i.e. the id is required and no partial updates are allowed.

• PATCH Updates/modifies a specific element. This method is idempotent, can only be called on a specific element, and replaces the given attributes with the data provided. For example PATCH https://example.com/api/contests/wf14/teams/10 with JSON contents {"name":"Our cool new team name"}.

• DELETE Delete a specific element. Idempotent, but will return a 404 error code when repeated. Any provided data is ignored, and there is no response body if successful. Example: DELETE https://example.com/api/contests/wf14/teams/8. Note that deletes must keep referential integrity intact.

Success, Failure, and HTTP Responses

Standard HTTP status codes are returned to indicate success or failure. If successful DELETE will have no response body, GET on a collection will return the collection, and every other method will contain the current (updated) state of the element.

If a POST, PUT, PATCH would cause any of the following issues it must fail, in addition to any endpoint or element-specific requirements:

• A PATCH on an id that doesn't exist. Will return a 404 error code.
• A PUT or PATCH containing an id that does not match the URL. Will return a 409 error code.
• A missing required attribute.
• An attribute that must not be provided is provided.
• An attribute type that is incorrect or otherwise invalid (e.g. non-nullable attribute set to null).
• A reference to another element is invalid (to maintain referential integrity).

In addition to any endpoint or element-specific requirements, DELETE must fail if the element id doesn't exist, and return a 404 error code. If the element being deleted is referenced by another element, the server must either fail or implement a cascading delete (to maintain referential integrity)

When there is a failure using any method the response message body must include a JSON element that contains the attributes ‘code' (a number, identical to the HTTP status code returned) and ‘message' (a string) with further information suitable for the user making the request, as per the following example:

{"code":403,
 "message":"Teams cannot send clarifications to another team"}

Roles

Access to this API is controlled via user roles. The API provider must require authentication to access each role except for optionally the public role. The API provider must support HTTP basic authentication (RFC). This provides a standard and flexible method; besides HTTP basic auth, other forms of authentication can be offered as well.

Each provider must support at least the following roles, although additional roles may be supported for specific uses:

• public (default role: contest data that's available to everyone)
• admin (data or capability only available to contest administrators)

Role-based access may completely hide some objects from the user, may omit certain attributes, or may embargo or omit objects based on the current contest time. By default, the public user has read-only access (no POST, PUT, PATCH or DELETE methods allowed) and does not have access to judgements and runs from submissions made after the contest freeze time.

Referential integrity

Some attributes in elements are references to IDs of other elements. When such an attribute has a non-null value, then the referenced element must exist. That is, the full set of data exposed by the API must at all times be referentially intact. This implies for example that before creating a team with an organization_id, the organization must already exist. In reverse, that organization can only be deleted after the team is deleted, or alternatively, the team's organization_id is set to null.

Furthermore, the ID attribute (see below) of elements are not allowed to change. However, note that a particular ID might be reused by first deleting an element and then creating a new element with the same ID.

JSON attribute types

Attribute types are specified as one of the standard JSON types, or one of the more specific types defined below. Implementations must be consistent with respect to the optional parts of each type, e.g. if the optional .uuu is included in any absolute timestamp it must be included when outputting all absolute timestamps.

• Strings (type string in the specification) are built-in JSON strings.
• Numbers (type number in the specification) are built-in JSON numbers.
• Booleans (type boolean in the specification) are built-in JSON booleans.
• Integers (type integer in the specification) are JSON numbers that are restricted to be integer. They should be represented in standard integer representation (-)?[0-9]+.
• Absolute timestamps (type TIME in the specification) are strings containing human-readable timestamps, given in ISO 8601 extended combined date/time format with timezone: yyyy-mm-ddThh:mm:ss(.uuu)?[+-]zz(:mm)? (or timezone Z for UTC).
• Relative times (type RELTIME in the specification) are strings containing human-readable time durations, given in a slight modification of the ISO 8601 extended time format: (-)?(h)*h:mm:ss(.uuu)?
• Identifiers (type ID in the specification) are given as string consisting of characters [a-zA-Z0-9_.-] of length at most 36 and not starting with a - (dash) or . (dot) or ending with a . (dot). IDs are unique within each endpoint. IDs are assigned by the person or system that is the source of the object, and must be maintained by downstream systems. For example, the person configuring a contest on disk will typically define the ID for each team, and any CCS or CDS that exposes the team must use the same ID. Some IDs are also used as identifiable labels and are marked below along with the recommended format. These IDs should be meaningful for human communication (e.g. team "43", problem "A") and are as short as reasonable but not more than 10 characters. IDs not marked as labels may be random characters and cannot be assumed to be suitable for display purposes.
• Ordinals (type ORDINAL in the specification) are used to give an explicit order to a list of objects. Ordinal attributes are integers and must be non-negative and unique in a list of objects, and they should typically be low numbers starting from zero. However, clients must not assume that the ordinals start at zero nor that they are sequential. Instead the ordinal values should be used to sort the list of objects.
• File references (types IMAGE, VIDEO, ARCHIVE and STREAM in the specification) are represented as a JSON object with elements as defined below.
• Arrays (type array of <type> in the specification) are built-in JSON arrays of some type defined above.

Element for file reference objects:

The href attributes may be absolute or relative URLs; relative URLs must be interpreted relative to the baseurl of the API. For example, if baseurl is https://example.com/api, then the following are equivalent JSON response snippets pointing to the same location:

  "href":"https://example.com/api/contests/wf14/submissions/187/files"
  "href":"contests/wf14/submissions/187/files"

For images, the supported mime types are image/png, image/jpeg, and image/svg+xml.

For images in SVG format, i.e. those having a mime type of image/svg+xml, the values of width and height should be the viewport width and height in pixels when possible, but otherwise the actual values don't matter as long as they are positive and represent the correct aspect ratio.

If implementing support for uploading files pointed to by resource links, substitute the href element with a data element with a base64 encoded string of the associated file contents as the value.

For example

PUT https://example.com/api/contests/wf14/organizations/inst105

with JSON data

{ "id":"inst105",
  "name":"Carnegie Mellon University",
  ...
  "logo": [{"data": "<base64 string>", "width": 160, "height": 160}]
}

Extensibility

This specification is meant to cover the basic data of contests, with the idea that server/client implementations can extend this with more data and/or roles. In particular, this specification already lists some endpoints or specific attributes as optional. The following guidelines are meant to ease extensibility.

• Clients should accept extra attributes in endpoints, that are not specified here.
• Servers should not expect clients to recognize more than the basic, required specification.
• In this specification and extensions, an attribute with value null may be left out by the server (i.e. not be present). A client must treat an attribute with value null equivalently as that attribute not being present.

Interface specification

The following list of API endpoints should be supported. Note that state, scoreboard and event-feed are singular nouns and indeed contain only a single element.

All endpoints should support GET; specific details on other methods are mentioned below.

Types of endpoints

The endpoints can be categorized into 3 groups as follows:

• Configuration: contests, judgement-types, languages, problems, groups, organizations, teams, team-members;
• Live data: state, submissions, judgements, runs, clarifications, awards, commentary;
• Aggregate data: scoreboard, event-feed.

Configuration is normally set before contest start. Is not expected to, but could occasionally be updated during a contest. It does not have associated timestamp/contest time attributes. Updates are notified via the event feed.

Live data is generated during the contest and new elements are expected. Data is immutable though: only inserts, no updates or deletes of elements. It does have associated timestamp/contest time attributes. Inserts and deletes are notified via the event feed. Note: judgements are the exception to immutability in a weak sense: they get updated once with the final verdict.

Aggregate data: Only GET makes sense. These are not included in the event feed, also note that these should not be considered proper REST endpoints, and that the event-feed endpoint is a streaming feed in NDJSON format.

Table column description

In the tables below, the columns are:

• Name: Attribute name; object sub-attributes are indicated as object.attribute.
• Type: Data type of the attribute; one of the types listed above.
• Required?: Whether this is a required attribute that must be implemented to conform to this specification.
• Nullable?: Whether the attribute might be null (and thus implicitly can also not be present in that case).
• Description: Description of the meaning of the attribute and any special considerations.

Note that attributes with null value may be left out by the server. Furthermore, optional attributes must still be consistently implemented (or not) *within* each contest. This implies the following for attributes that are:

• Required and not nullable: The attribute must always be present with a value.
• Required and nullable: The attribute may be null, and only in that case it may be left out.
• Optional and not nullable: The attribute may not be implemented, but that implies that no element of the endpoint has the attribute set. If one element has this attribute present, then it must be not null and the same must be true for all same type elements within the contest.
• Optional and nullable: The attribute may be null or not present. In the latter case that can either be because it was a left out null value or because it was not implemented.

Filtering

Endpoints that return a JSON array must allow filtering on any attribute with type ID (except the id field) by passing it as a query argument. For example, clarifications can be filtered on the recipient by passing to_team_id=X. To filter on a null value, pass an empty string, i.e. to_team_id=. It must be possible to filter on multiple different fields simultaneously, with the meaning that all conditions must be met (they are logically ANDed). Note that filtering on any other field, including fields with the type array of ID, does not have to be supported.

Contests

Provides information on the current contest.

The following endpoint is associated with contest:

JSON elements of contest objects:

The expected/typical use of countdown_pause_time is that once a start_time is defined and close, the countdown may be paused due to unforeseen delays. In this case, start_time should be set to null and countdown_pause_time to the number of seconds left to count down. The countdown_pause_time may change to indicate approximate delay. Countdown is resumed by setting a new start_time and resetting countdown_pause_time to null.

PATCH start_time

Implementations must have a role that has the ability to clear or set the contest start time via a PATCH method.

The PATCH must include a valid JSON element with only two or three attributes allowed: the contest id (used for verification), a start_time (a <TIME> value or null), and an optional countdown_pause_time (<RELTIME>). As above, countdown_pause_time can only be non-null when start time is null.

The request should fail with a 401 error code if the user does not have sufficient access rights, or a 403 error code if the contest is started or within 30s of starting, or if the new start time is in the past or within 30s.

Examples

Request:

GET https://example.com/api/contests/wf2014

Returned data:

{
   "id": "wf2014",
   "name": "2014 ICPC World Finals",
   "formal_name": "38th Annual World Finals of the ACM International Collegiate Programming Contest",
   "start_time": "2014-06-25T10:00:00+01",
   "duration": "5:00:00",
   "scoreboard_freeze_duration": "1:00:00",
   "scoreboard_type": "pass-fail",
   "penalty_time": 20,
   "banner": [{
       "href": "https://example.com/api/contests/wf2014/banner",
       "width": 1920,
       "height": 240
   }]
}

Request:

GET https://example.com/api/contests/dress2016

Returned data:

{
   "id": "dress2016",
   "name": "2016 ICPC World Finals Dress Rehearsal",
   "start_time": null,
   "countdown_pause_time": "0:03:38.749",
   "duration": "2:30:00"
}

Request:

PATCH https://example.com/api/contests/wf2014

Request data:

{
   "id": "wf2014",
   "start_time": "2014-06-25T10:00:00+01"
}

Request:

PATCH https://example.com/api/contests/wf2016

Request data:

{
   "id": "wf2016",
   "start_time": null
}

Judgement Types

Judgement types are the possible responses from the system when judging a submission.

The following endpoints are associated with judgement types:

JSON elements of judgement type objects:

Known judgement types

The list below contains standardized identifiers for known judgement types. These identifiers should be used by a server. Please create a pull request at https://github.com/icpc/ccs-specs when there are judgement types missing.

The column Big 5 lists the "big 5" equivalents, if any. A * in the column means that the judgement is one of the "big 5".

The Translation column lists other judgements the judgement can safely be translated to, if a system does not support it.

Examples

Request:

GET https://example.com/api/contests/wf14/judgement-types

Returned data:

[{
   "id": "CE",
   "name": "Compiler Error",
   "penalty": false,
   "solved": false
}, {
   "id": "AC",
   "name": "Accepted",
   "penalty": false,
   "solved": true
}]

Request:

GET https://example.com/api/contests/wf14/judgement-types/AC

Returned data:

{
   "id": "AC",
   "name": "Accepted",
   "penalty": false,
   "solved": true
}

Languages

Languages that are available for submission at the contest.

The following endpoints are associated with languages:

JSON elements of language objects:

JSON elements of Command objects:

The compiler and runner elements are intended for informational purposes. It is not expected that systems will synchronize compiler and runner settings via this interface.

Known languages

Below is a list of standardized identifiers for known languages with their name, extensions and entry point name (if any). When providing one of these languages, the corresponding identifier should be used. The language name and entry point name may be adapted e.g. for localization or to indicate a particular version of the language. In case multiple versions of a language are provided, those must have separate, unique identifiers. It is recommended to choose new identifiers with a suffix appended to an existing one. For example cpp17 to specify the ISO 2017 version of C++.

Examples

Request:

GET https://example.com/api/contests/wf14/languages

Returned data:

[{
   "id": "java",
   "name": "Java",
   "entry_point_required": true,
   "entry_point_name": "Main class",
   "extensions": ["java"],
   "compiler": {
      "command": "javac",
      "args": "-O {files}",
      "version": "javac 11.0.4",
      "version-command": "javac --version"
   },
   "runner": {
      "command": "java",
      "version": "openjdk version \"11.0.4\" 2019-07-16"
   }
}, {
   "id": "cpp",
   "name": "GNU C++",
   "compiler": {
      "command": "gcc",
      "args": "-O2 -Wall -o a.out -static {files}",
      "version": "gcc (Ubuntu 8.3.0-6ubuntu1) 8.3.0"
   },
   "entry_point_required": false,
   "extensions": ["cc", "cpp", "cxx", "c++", "C"],
}, {
   "id": "python3",
   "name": "Python 3",
   "entry_point_required": true,
   "entry_point_name": "Main file",
   "extensions": ["py"]
}]

Problems

The problems to be solved in the contest

The following endpoints are associated with problems:

JSON elements of problem objects:

Examples

Request:

GET https://example.com/api/contests/wf14/problems

Returned data:

[{"id":"asteroids","label":"A","name":"Asteroid Rangers","ordinal":1,"color":"blue","rgb":"#00f","time_limit":2,"test_data_count":10},
 {"id":"bottles","label":"B","name":"Curvy Little Bottles","ordinal":2,"color":"gray","rgb":"#808080","time_limit":3.5,"test_data_count":15}
]

Request:

GET https://example.com/api/contests/wf14/problems/asteroids

Returned data:

{"id":"asteroids","label":"A","name":"Asteroid Rangers","ordinal":1,"color":"blue","rgb":"#00f","time_limit":2,"test_data_count":10}

Groups

Grouping of teams. At the World Finals these are the super regions; at other contests these may be the different sites, divisions, or types of contestants.

Teams may belong to multiple groups. For instance, there may be a group for each site, a group for university teams, a group for corporate teams, and a group for ICPC-eligible teams. Teams could belong to two or three of these. When there are different kinds of groups for different purposes (e.g. sites vs divisions), each group or set of groups should have a different type attribute (e.g. "type":"site" and "type":"division").

Groups must exist for any combination of teams that must be ranked on a group scoreboard, which means groups may be created for combinations of other groups. For instance, if there is a requirement to show a scoreboard for teams in each of D divisions at every one of S sites, then in addition to the D + S groups there will also be DxS combined/product groups. It is recommended that these groups have a type like "type":"<group1>-<group2>", e.g. "type":"site-division".

The following endpoints are associated with groups:

Note that these endpoints must be provided if groups are used. If they are not provided no other endpoint may refer to groups (i.e. return any group_ids).

JSON elements of group objects:

Known group types

The list below contains standardized identifiers for known group types. These identifiers should be used when the purpose of a group matches.

Examples

Request:

GET https://example.com/api/contests/wf14/groups

Returned data:

[
  {"id":"asia-74324325532","icpc_id":"7593","name":"Asia"}
]

Request:

GET https://example.com/api/contests/wf14/groups

Returned data:

[
  {"id":"42425","name":"Division 2","type":"division"}
]

Organizations

Teams can be associated with organizations which will have some associated information, e.g. a logo. Typically organizations will be universities.

The following endpoints are associated with organizations:

Note that the first two endpoints must be provided if organizations are used. If they are not provided no other endpoint may refer to organizations (i.e. return any organization_ids).

JSON elements of organization objects:

Examples

Request:

GET https://example.com/api/contests/<id>/organizations

Returned data:

[{"id":"inst123","icpc_id":"433","name":"Shanghai Jiao Tong U.","formal_name":"Shanghai Jiao Tong University"},
 {"id":"inst105","name":"Carnegie Mellon University","country":"USA",
  "logo":[{"href":"http://example.com/api/contests/wf14/organizations/inst105/logo/56px","width":56,"height":56},
          {"href":"http://example.com/api/contests/wf14/organizations/inst105/logo/160px","width":160,"height":160}]
 }
]

Teams

Teams competing in the contest.

The following endpoints are associated with teams:

JSON elements of team objects:

Examples

Request:

GET https://example.com/api/contests/wf14/teams

Returned data:

[{"id":"11","icpc_id":"201433","name":"Shanghai Tigers","organization_id":"inst123","group_ids":["asia-74324325532"]},
 {"id":"123","name":"CMU1","organization_id":"inst105","group_ids":["8","11"]}
]

Team members

Team members of teams in the contest.

The following endpoints are associated with languages:

JSON elements of team member objects:

Examples

Request:

GET https://example.com/api/contests/wf14/team-members

Returned data:

[{"id":"john-smith","team_id":"43","icpc_id":"32442","first_name":"John","last_name":"Smith","sex":"male","role":"contestant"},
 {"id":"osten-umlautsen","team_id":"43","icpc_id":null,"first_name":"Östen","last_name":"Ümlautsen","sex":null,"role":"coach"}
]

Contest state

Current state of the contest, specifying whether it's running, the scoreboard is frozen or results are final.

The following endpoints are associated with state:

JSON elements of state objects:

These state changes must occur in the order listed in the table above, as far as they do occur, except that thawed and finalized may occur in any order. For example, the contest may never be frozen and hence not thawed either, or, it may be finalized before it is thawed. That, is the following sequence of inequalities must hold:

started < frozen < ended < thawed    < end_of_updates,
                   ended < finalized < end_of_updates.

A contest that has ended, has been thawed (or was never frozen) and is finalized must not change. Thus, end_of_updates can be set once both finalized is set and thawed is set if the contest was frozen.

Examples

Request:

GET https://example.com/api/contests/wf14/state

Returned data:

{
  "started": "2014-06-25T10:00:00+01",
  "ended": null,
  "frozen": "2014-06-25T14:00:00+01",
  "thawed": null,
  "finalized": null,
  "end_of_updates": null
}

Submissions

Submissions, a.k.a. attempts to solve problems in the contest.

The following endpoints are associated with submissions:

JSON elements of submission objects:

The entry_point attribute must be included for submissions in languages which do not have a single, unambiguous entry point to run the code. In general the entry point is the string that needs to be specified to point to the code to be executed. Specifically, for Python it is the file name that should be run, and for Java and Kotlin it is the fully qualified class name (that is, with any package name included, e.g. com.example.myclass for a class in the package com.example in Java). For C and C++ no entry point is required and it must therefore be null.

The files attribute provides the file(s) of a given submission as a zip archive. These must be stored directly from the root of the zip file, i.e. there must not be extra directories (or files) added unless these are explicitly part of the submission content.

POST and PUT submissions

To add a submission one can use the POST method on the submissions endpoint or the PUT method directly on an element url. POST is typically used by teams submitting to the contest and PUT is used by admin users or tools. Both must include a valid JSON object with the same attributes the submission endpoint returns with a GET request with the following exceptions:

• The attributes team_id, time, and contest_time are optional depending on the use case (see below). The server may require attributes to either be absent or present, and should respond with a 4xx error code in such cases.
• Since files only supports application/zip, providing the mime field is optional.
• reaction may be provided but a CCS does not have to honour it.
• If the CCS supports a team role they will only have access to POST, and time must not be provided when using this role. time will always be set to the current time as determined by the server. team_id may be provided but then must match the ID of the team associated with the request.
• Simple proxies using an admin role should also use POST to let the server determine the id. team_id must be provided but time must not and will be determined by the server.
• For more advanced scenarios an admin role may use a PUT. time is required and the client is responsible for making sure the provided id is unique. For example in a setup with a central CCS with satellite sites where teams submit to a proxy CCS that forwards to the central CCS, this might be useful to make sure that the proxy CCS can accept submissions even when the connection to the central CCS is down. The proxy can then forward these submissions later, when the connection is restored again.

The request must fail with a 4xx error code if any of the following happens:

• A required attribute is missing.
• An attribute that must not be provided is provided.
• The supplied problem, team or language can not be found.
• An entrypoint is required for the given language, but not supplied.
• The mime field in files is set but invalid.
• Something is wrong with the submission file. For example it contains too many files, it is too big, etc.
• The provided id already exists or is otherwise not acceptable.

The response will contain a Location header pointing to the newly created submission and the response body will contain the initial state of the submission.

Performing a POST or PUT by any roles other than admin and team is not supported.

Use cases for POSTing and PUTting submissions

The POST and PUT submissions endpoint can be used for a variety of reasons, and depending on the use case, the server might require different fields to be present. A number of common scenarios are described here for informational purposes only.

Team client submitting to CCS

The most obvious and probably most common case is where a team directly submits to the CCS, e.g. with a command-line submit client.

In this case the client has the team role and a specific team_id already associated with it. POST must be used and the attributes id, team_id, time, and contest_time should not be specified; the server will determine these attributes and should reject submissions specifying them, or may ignore a team_id that is identical to the one that the client has authenticated as.

A proxy server forwarding to a CCS

A proxy server may receive submissions from team clients (like above) and forward these to a CCS. This might be useful, for example, in a multi-site contest setup, where each site runs a proxy that would still be reachable if connectivity with the central CCS is lost, or where the proxy forwards the submission to multiple CCS's that run in parallel (like the shadowing setup at the ICPC World Finals).

In such a scenario, the proxy server would timestamp the submissions and authenticate the submitting team, and then forward the submission to the upstream CCS using the admin role. The proxy would provide team_id and time attributes and the CCS should then accept and use these.

To allow the proxy to return a submission id during connectivity loss, each site could be assigned a unique prefix such that the proxy server itself can generate unique ids and then submit a PUT to the central CCS with the id attribute included. The central CCS should then accept and use that id attribute.

Further potential extensions

To allow for any further use cases, the specification is deliberately flexible in how the server can handle optional attributes.

• The contest_time attribute should normally not be specified when time is already specified as it can be calculated from time and the wallclock time is unambiguously defined without reference to contest start time. However, in a case where one would want to support a multi-site contest where the sites run out of sync, the use of contest_time might be considered.

Examples

Request:

GET https://example.com/api/contests/wf14/submissions

Returned data:

[{"id":"187","team_id":"123","problem_id":"10-asteroids",
  "language_id":"1-java","time":"2014-06-25T11:22:05.034+01","contest_time":"1:22:05.034","entry_point":"Main",
  "files":[{"href":"contests/wf14/submissions/187/files","mime":"application/zip"}]}
]

Note that the relative link for files points to the location https://example.com/api/contests/wf14/submissions/187/files since the base URL for the API is https://example.com/api.

Request:

POST https://example.com/api/contests/wf14/submissions

Request data:

{
   "language_id":"1-java",
   "problem_id":"10-asteroids",
   "team_id":"123",
   "time":"2014-06-25T11:22:05.034+01",
   "entry_point":"Main",
   "files":[{"data": "<base64 string>"}]
}

Returned data:

{
   "id":"187",
   "language_id":"1-java",
   "problem_id":"10-asteroids",
   "team_id":"123",
   "time":"2014-06-25T11:22:05.034+01",
   "contest_time":"1:22:05.034",
   "entry_point":"Main",
   "files":[{"href":"contests/wf14/submissions/187/files","mime":"application/zip"}]
}

Judgements

Judgements for submissions in the contest.

The following endpoints are associated with judgements:

JSON elements of judgement objects:

When a judgement is started, each of judgement_type_id, end_time and end_contest_time will be null (or missing). These are set when the judgement is completed.

Examples

Request:

GET https://example.com/api/contests/wf14/judgements

Returned data:

[{"id":"189549","submission_id":"wf2017-32163123xz3132yy","judgement_type_id":"CE","start_time":"2014-06-25T11:22:48.427+01",
  "start_contest_time":"1:22:48.427","end_time":"2014-06-25T11:23:32.481+01","end_contest_time":"1:23:32.481"},
 {"id":"189550","submission_id":"wf2017-32163123xz3133ub","judgement_type_id":null,"start_time":"2014-06-25T11:24:03.921+01",
  "start_contest_time":"1:24:03.921","end_time":null,"end_contest_time":null}
]

Runs

Runs are judgements of individual test cases of a submission.

The following endpoints are associated with runs:

JSON elements of run objects:

Examples

Request:

GET https://example.com/api/contests/wf14/runs

Returned data:

[{"id":"1312","judgement_id":"189549","ordinal":28,"judgement_type_id":"TLE",
  "time":"2014-06-25T11:22:42.420+01","contest_time":"1:22:42.420"}
]

Clarifications

Clarification message sent between teams and judges, a.k.a. clarification requests (questions from teams) and clarifications (answers from judges).

The following endpoints are associated with clarification messages:

JSON elements of clarification message objects:

Note that at least one of from_team_id and to_team_id has to be null. That is, teams cannot send messages to other teams.

POST and PUT clarifications

To add a clarification one can use the POST or PUT method on the clarifications endpoint. The POST or PUT must include a valid JSON object with the same attributes the clarification endpoint returns with a GET request with the following exceptions:

• When an attribute value would be null it is optional - you do not need to include it. e.g. if a clarification is not related to a problem you can chose to include or exclude the problem_id.
• When submitting using a team role, POST must be used and id, to_team_id, time, and contest_time must not be provided. from_team_id may be provided but then must match the ID of the team associated with the request. The server will determine an id and the current time and contest_time.
• When submitting using an admin role, POST or PUT may be used and id, time, and contest_time may be required to either be absent or present depending on the use case, e.g. whether the server is the CCS, is acting as a proxy, or a caching proxy. See notes under the submission interface for more detail. In cases where these attributes are not allowed the server will respond with a 4xx error code.

The request must fail with a 4xx error code if any of the following happens:

• A required attribute is missing.
• An attribute that must not be provided is provided.
• The supplied problem, from_team, to_team, or reply_to cannot be found or are not visible to the role that's submitting.
• The provided id already exists or is otherwise not acceptable.

The response will contain a Location header pointing to the newly created clarification and the response body will contain the initial state of the clarification.

Performing a POST or PUT by any roles other than admin and team is not supported.

Examples

Request:

GET https://example.com/api/contests/wf14/clarifications

Returned data:

[{"id":"wf2017-1","from_team_id":null,"to_team_id":null,"reply_to_id":null,"problem_id":null,
  "text":"Do not touch anything before the contest starts!","time":"2014-06-25T11:59:27.543+01","contest_time":"-0:15:32.457"}
]

Request:

GET https://example.com/api/contests/wf14/clarifications

Returned data:

[{"id":"1","from_team_id":"34","to_team_id":null,"reply_to_id":null,"problem_id":null,
  "text":"May I ask a question?","time":"2017-06-25T11:59:27.543+01","contest_time":"1:59:27.543"},
 {"id":"2","from_team_id":null,"to_team_id":"34","reply_to_id":"1","problem_id":null,
  "text":"Yes you may!","time":"2017-06-25T11:59:47.543+01","contest_time":"1:59:47.543"}
]

Request:

GET https://example.com/api/contests/wf14/clarifications

Returned data:

[{"id":"1","from_team_id":"34","text":"May I ask a question?","time":"2017-06-25T11:59:27.543+01","contest_time":"1:59:27.543"},
 {"id":"2","to_team_id":"34","reply_to_id":"1","text":"Yes you may!","time":"2017-06-25T11:59:47.543+01","contest_time":"1:59:47.543"}
]

Request:

POST https://example.com/api/contests/wf14/clarifications

Request data:

{
   "problem_id":"10-asteroids",
   "from_team_id":"34",
   "text":"Can I assume the asteroids are round?"
}

Returned data:

{
   "id":"clar-43",
   "problem_id":"10-asteroids",
   "from_team_id":"34",
   "text":"Can I assume the asteroids are round?",
   "time":"2017-06-25T11:59:47.543+01",
   "contest_time":"1:59:47.543"
}

Awards

Awards such as medals, first to solve, etc.

The following endpoints are associated with awards:

JSON elements of award objects:

Semantics

• Awards are not final until the contest is.
• An award may be created at any time, although it is recommended that a system creates the awards it intends to award before the contest starts.
• If an award has a non-null team_ids, then it must be kept up to date during the contest. E.g. if "winner" will not be updated with the current leader during the contest, it must be null until the award is updated.
• If an award is present during the contest this means that if the contest would end immediately and then become final, that award would be final. E.g. the "winner" during the contest should be the current leader. This is of course subject to what data the client can see; the public role's winner may not change during the scoreboard freeze but an admin could see the true current winner.

Known awards

For some common award cases the following IDs should be used.

POST, PUT, PATCH, and DELETE awards

Clients with the admin role may make changes to awards using the normal HTTP methods as specified above. Specifically, they can POST new awards, create or replace one with a known id via PUT, PATCH one or more attributes, or DELETE an existing award.

The server may be configured to manage (assign or update) some award ids, and may block clients from modifying them. However, if a client is able to modify an award it must assume that it is responsible for managing that award id unless and until it sees an indication that something else is now managing that award - either a change that it did not request, or a future modification fails.

For example, the server may be configured to assign the winner award and not allow any client to modify it. The same server may assign *-medal awards by default, but allow clients to modify them. Once a client modifies any of the *-medal awards, it is responsible for updating it if anything changes. Likewise, the client could add any arbitrary awards like first-submission-for-country-* and would be responsible for managing these.

The server should create all the awards it is configured to manage before the start of the contest, so that clients can know which awards are already handled.

The request must fail with a 4xx error code if any of the following happens:

• A POST that includes an id.
• A PATCH, or DELETE on an award that doesn't exist.
• A POST or PUT that is missing one of the required attributes (citation and team_ids).
• A PATCH that contains an invalid attribute (e.g. null citation or team_ids).
• A PUT or PATCH that includes an award id that doesn't match the id in the url.
• A POST, PUT, PATCH, or DELETE on an award id that the server is configured to manage exclusively.

Examples

Request:

GET https://example.com/api/contests/wf14/awards

Returned data:

[{"id":"gold-medal","citation":"Gold medal winner","team_ids":["54","23","1","45"]},
 {"id":"first-to-solve-a","citation":"First to solve problem A","team_ids":["45"]},
 {"id":"first-to-solve-b","citation":"First to solve problem B","team_ids":[]}
]

Request:

POST https://example.com/api/contests/wf14/awards

Request data:

{"citation":"Best team costumes","team_ids":["42"]}

Response data:

{"id":"best-costume","citation":"Best team costumes","team_ids":["42"]}

Request:

PUT https://example.com/api/contests/wf14/awards/best-costume

Request data:

{"id":"best-costume","citation":"Best team costumes","team_ids":["24"]}

Request:

PATCH https://example.com/api/contests/wf14/awards/best-costume

Request data:

{"citation":"Best team cosplay"}

Request:

DELETE https://example.com/api/contests/wf14/awards/best-costume

Commentary

Commentary on events happening in the contest

The following endpoints are associated with commentary:

JSON elements of award objects:

For the message, if an literal # is needed, \# must be used. Similarly for literal \, \\ must be used.

Examples

Request:

GET https://example.com/api/contests/wf14/commentary

Returned data:

[{"id":"143730", "time":"2021-03-06T19:02:02.328+00", "contest_time":"0:02:02.328", "message": "#t20 made a submission for #panttyping. If correct, they will solve the first problem and take the lead", "team_ids": ["314089"], "problem_ids": ["anttyping"]}, 
 {"id": "143736", "time": "2021-03-06T19:02:10.858+00", "contest_time": "0:02:10.858", "message": "#t20 fails its first attempt on #panttyping due to WA", "team_ids": ["314089"], "problem_ids": ["anttyping"]}, 
 {"id": "143764", "time": "2021-03-06T19:03:07.517+00", "contest_time": "0:03:07.517", "message": "#t24 made a submission for #pmarch6. If correct, they will solve the first problem and take the lead", "team_ids": ["314115"], "problem_ids": ["magictrick"]}
]

Scoreboard

Scoreboard of the contest.

Since this is generated data, only the GET method is allowed here, irrespective of role.

The following endpoint is associated with the scoreboard:

Scoreboard request options

The following options can be passed to the scoreboard endpoint.

Scoreboard at the time of a given event

By passing an event ID with the after_event_id URL argument, the scoreboard can be requested as it was directly after the specified event:

/scoreboard?after_event_id=xy1234

This makes it possible for a client to get the scoreboard information that is guaranteed to match a certain contest event. In case no after_event_id argument is provided, the current scoreboard will be returned. The request will fail with a 400 error if the id is invalid.

A suggested efficient server-side implementation to provide this, is to store with each event that changes the scoreboard, the new team scoreboard row.

Group scoreboard

By passing group_id with a valid group ID a scoreboard can be requested for the teams in a particular group:

/scoreboard?group_id=site1

Each group scoreboard is ranked independently and contains only the teams that belong to the specified group. If a client wants to know ‘local' vs ‘global' rank it can query both the group and primary scoreboards.

A 4xx error code will be returned if the group id is not valid. Groups that are not included in the groups endpoint for the role making the request are not valid.

Scoreboard format

JSON elements of the scoreboard object.

The scoreboard rows array is sorted according to rank and alphabetical on team name within identically ranked teams. Here alphabetical ordering means according to the Unicode Collation Algorithm, by default using the en-US locale.

Each JSON object in the rows array consists of:

Each problem object within the scoreboard consists of:

Examples

Request:

GET https://example.com/api/contests/wf14/scoreboard

Returned data:

{
  "event_id": "xy1234",
  "time": "2014-06-25T14:13:07.832+01",
  "contest_time": "4:13:07.832",
  "state": {
    "started": "2014-06-25T10:00:00+01",
    "ended": null,
    "frozen": "2014-06-25T14:00:00+01",
    "thawed": null,
    "finalized": null,
    "end_of_updates": null
  },
  "rows": [
    {"rank":1,"team_id":"123","score":{"num_solved":3,"total_time":340},"problems":[
      {"problem_id":"1","num_judged":3,"num_pending":1,"solved":false},
      {"problem_id":"2","num_judged":1,"num_pending":0,"solved":true,"time":20},
      {"problem_id":"3","num_judged":2,"num_pending":0,"solved":true,"time":55},
      {"problem_id":"4","num_judged":0,"num_pending":0,"solved":false},
      {"problem_id":"5","num_judged":3,"num_pending":0,"solved":true,"time":205}
    ]}
  ]
}

Event feed

Provides the event (notification) feed for the current contest. This is effectively a changelog of create, update, or delete events that have occurred in the REST endpoints. Some endpoints (specifically the Scoreboard and the Event feed itself) are aggregated data, and so these will only ever update due to some other REST endpoint updating. For this reason there is no explicit event for these, since there will always be another event sent. This can also be seen by the fact that there is no scoreboard event in the table of events below.

Every notification provides the current state of a single contest object. There is no guarantee on order of events (except for general requirements below), whether two consecutive changes cause one or two events, duplicate events, or even that different clients will receive the same order or set of events. The only guarantee is that when an object changes one or more times you'll receive an event, and the latest event received for any object is the correct and current state of that object (e.g. if an object was created and deleted you'll always receive a delete event last).

As a concrete example, judgement events are usually fired when judging is started, and fired again when the final judgement is available. If a client connects after the judgement, or a client was disconnected during the judgement, they will typically only receive the final (complete) judgement.

There are two mechanisms that clients can use to receive events: a webhook, or a streaming HTTP feed. Both mechanisms have the same format (payload) and events, but different benefits, drawbacks, and ways to register. Webhooks are better for internet-scale, asynchronous processing, and disconnected systems; the HTTP feed is better for browser-based applications and onsite contests.

General requirements

The event responses and data objects contained in it must observe the same restrictions as those of the respective endpoints they represent. This means that attributes inside the data element will be present if and only if the client has access to those at the respective endpoint. The client only receives create, update and delete events of elements it has (partial) access to. When time-based access is granted or revoked, create or delete events are dispatched for each affected entity.

Referential integrity must be strictly adhered to for new objects. i.e. if there is a new object that refers to another object (e.g. a submission for a team) then the referenced object must already exist and have been notified. There is no such guarantee for deletion: if an object that is referred to is deleted then by definition all of the child objects will be deleted, but the events may not arrive in strict referential order:

• If an object, A, refers to another object, B, then the event that shows that A has been created or updated to refer to B must come after the event that shows that B has been created.
• If some data is only available after a specific state change, then the event showing the state change must come before any update events making that data available. E.g. problems are only available after contest start for the public role, so the state event showing that the contest has started must come before the problem events creating the problems.
• Since nothing must change after the contest has ended, thawed (or never been frozen), and been finalized, only the end_of_updates event may come after the state event showing that.

Feed format

The feed is served as JSON objects, with every event corresponding to a change in a single object (submission, judgement, language, team, etc.) or full endpoint. The general format for events is:

{"contest_id": "<id>", "endpoint": "<endpoint>", "id": "<id>", "data": <JSON data for element> }

All event types have a corresponding API endpoint, as specified in the table below.

Note that this does not contain /webhooks/<id>, since no events will be triggered for that endpoint.

Filtering

TODO - filter by contest id and/or endpoint

Examples

The following are examples of contest events:

{"contest_id":"finals","endpoint":"problems","id":null,"data":[
   {"id":"asteroids","label":"A","name":"Asteroid Rangers","ordinal":1,"color":"blue","rgb":"#00f","time_limit":2,"test_data_count":10},
   {"id":"bottles","label":"B","name":"Curvy Little Bottles","ordinal":2,"color":"gray","rgb":"#808080","time_limit":3.5,"test_data_count":15}]}

{"contest_id":"finals","endpoint":"state","id":null,"data":{
   "started": "2014-06-25T10:00:00+01",
   "ended": null,
   "frozen": "2014-06-25T14:00:00+01",
   "thawed": null,
   "finalized": null,
   "end_of_updates": null}}

{"contest_id":"finals","endpoint":"teams","id":"11","data":{"id":"11","icpc_id":"201433","name":"Shanghai Tigers","organization_id":"inst123","group_id":"asia"}}

{"contest_id":"finals","endpoint":"teams","id":"11","data":{"id":"11","icpc_id":"201433","name":"The Shanghai Tigers","organization_id":"inst123","group_id":"asia"}}

{"contest_id":"finals","endpoint":"teams","id":"11","data":null}

TODO: data is object or array - is that too ugly?

Webhook

A webhook allows you to receive HTTP callbacks whenever there is a change to the contest. Clients are only notified of future changes; they are expected to use other mechanisms if they need to determine the current state of the contest. Every callback will contain one JSON object as specified above.

Responding to each event with a 2xx response code indicates successful receipt and ensures that the events in the payload are never sent again. If the client responds with anything other than 2xx, the server will continue to periodically try again, potentially with different payloads (e.g. as new events accumulate). Callbacks to each client are always sent synchronously and in order; clients do not need to worry about getting callbacks out of order and should always process each callback fully before processing the next one.

If the client fails to respond to multiple requests over a period of time (configured for each contest), it will be assumed deactivated and automatically removed from future callbacks.

The following endpoints is associated with the webhook:

JSON elements of webhook callback objects:

Adding a webhook

To register a webhook, you need to post your server's callback URL. To do so, perform a POST request with a JSON body with the fields (except id) from the above table to the /webhooks endpoint together with one additional field, called token. In this field put a client-generated token that can be used to verify that callbacks come from the CCS. If you don't supply contest_ids and/or endpoints, they will default to [].

Examples

Request:

POST https://example.com/api/webhooks

Payload:

{"url": "https://myurl", "token": "mysecrettoken" }

Request:

GET https://example.com/api/webhooks

Returned data:

[{
    "id":"icpc-live",
    "url":"https://myurl",
    "endpoints": [],
    "contest_ids": [],
    "active": true
},{
    "id":"shadow",
    "url":"https://myotherurl",
    "endpoints": ["teams", "problems"],
    "contest_ids": ["wf2014"],
    "active": false
}]

When the CCS wants to send out a callback, it will check all active webhooks, filter them on applicable endpoint and contest ID and perform a POST to the URL. The CCS will add a header to this request called Webhook-Token which contains the token as supplied when creating the webhook. Clients should verify that this token matches with what they expect. The body of the request will be in the same format as in the feed format, i.e. it contains the keys contest_id, endpoint, id and data.

HTTP Feed

The HTTP event feed is a streaming HTTP endpoint that allows connected clients to receive contest events. The feed is a complete log of contest objects that starts "at the beginning of time" so all existing objects will be sent upon initial connection, but apart from referential integrity requirements they may appear in any order (e.g. teams or problems first).

Each line is an NDJSON formatted contest event as specified above. The feed does not terminate under normal circumstances, so to ensure keep alive a newline must be sent if there has been no event within 120 seconds.

Since this is generated data, only the GET method is allowed here, irrespective of role.

The following endpoint is associated with the event feed:

Reconnection

If a client loses connection or needs to reconnect after a brief disconnect (e.g. client restart), it can use the ‘time' argument to specify the last event it received:

/event-feed?time=xx

If specified, the server will attempt to start sending events around the given time to reduce the volume of events and required reconciliation. If the time passed is too large or the server does not support this attribute, all objects will be sent. There is no guarantee that all updates (e.g. a team name correction, which is not time-based) that occurred during the time the client was disconnected will be reflected.