Algorithms

Userfeeds Platform allows running community written algorithms on too of data gathered from all supported sources like Ethereum blockchain (mainnet, ropsten, rinkeby, kovan, whisper messages), IPDB (BigchainDB) or direct messages.

Algorithms can be referenced by their identifier that depends on where the author decided to share them.

Some of algorithm sources and naming schemes are:

• Github - github:username.repository.filename
• Bitbucket - bitbucket:username.repository.filename
• Gist - gist:hash

Note

More esoteric sources are possible, like downloading code from blockchain or encrypted algorithm code but their support is not planned at the moment.

Warning

Current support is limited to built-in algorithms through their short identifiers.

Built-in algorithms

Source code: https://github.io/Userfeeds/algorithms

All algorithms require one information: context. They will get all claims intended for given context and sort them according to their specification.

Note

Built-in algorithms can be referenced normally via eg. github:username.repo.filename or by their short name eg. latest

Claims (TODO)

Identifier: claims, github:Userfeeds.algorithms.claims

The simples algorithm of all. It just returns 100 latest claims for given context

Parameters

type

• string
• optional

Type of claims to return

Example data

{
  "items": [
    {
      "created_at": 1499676056549,
      "id": "0xd10409e340b5f41be0c0c8db85a61b2db0e5f4fb6c87fd3a5e4b73adc9bee9bf",
      "score": 1000000000000000000,
      "target": "https://userfeeds.io/"
    },
    {
      "created_at": 1499676053781,
      "id": "0xcd05844f64d472d8aeceeafc1c14f377bc664a9f94574ee820128254c962bfdc",
      "score": 5000000000000000000,
      "target": "https://example.com/"
    },
    {
      "created_at": 1499676051855,
      "id": "0xaa6b523d72d1f0e9b00689f494ac622804e0552e55c939cabc97381bd401732b",
      "score": 0,
      "target": "0x64537b9f7c9d85bae5c52cb4fa2307e7da14c6f1cae6d710b66a4ddcc059e1a8"
    },
    {
      "created_at": 1499676051757,
      "id": "0x64537b9f7c9d85bae5c52cb4fa2307e7da14c6f1cae6d710b66a4ddcc059e1a8",
      "score": 100000000000000,
      "target": "https://some.path/"
    }
  ]
}

Authored (TODO)

Identifier: authored, github:Userfeeds.algorithms.authored

This algorithm return all claims that were authored by context. In other words all claims that have AUTHORED relation to Account node with id equal to context.

Parameters

type

• string
• optional

Type of claims to return

Example data

{
  "items": [
    {
      "created_at": 1499676051855,
      "id": "0xaa6b523d72d1f0e9b00689f494ac622804e0552e55c939cabc97381bd401732b",
      "score": 0,
      "target": "0x64537b9f7c9d85bae5c52cb4fa2307e7da14c6f1cae6d710b66a4ddcc059e1a8"
    },
    {
      "created_at": 1499675901184,
      "id": "0xc3921abdd00ca53925e5a5b8559537d44e19165ff3ad4e89daf0bb034a4afa21",
      "score": 10000000000000000,
      "target": "Lorem ipsum dolor sit amet, consectetur adipisicing elit"
    },
    {
      "created_at": 1499156515246,
      "id": "0x85b6643f69e746c7525fc3ed88e1f18241068bcff6516fbf2dd0d967ebdaf390",
      "score": 0,
      "target": "0x794d0c42fc35af7eba9216261ea3799e19d079d24dcf795fda94acf20ad298f8"
    }
  ]
}

Sponsored

Identifier: sponsored, github:Userfeeds.algorithms.sponsored

This algorithm is different form other as it does not use claims as a source of information but more low level transaction.input and as such works only for contexts like ethereum.

It just takes whatever is inside transaction.input sums up all transaction.value, sorts the result and returns those inputs as target.

The idea is that developer would send inside transaction.input only values that make sense for their application and only this application can display appropriate content for each entry.

An example application can be a gallery of images with identifiers like img1 img2... Using those identifiers as transaction.input values, developer can create special sponsored gallery in his application that will display images sorted by overall value of ether used to promote them.

Example application can be found at Quick Start.

Example data

{
  "items": [
    {
      "bids": 2,
      "score": 1.01,
      "target": "0x636174"
    },
    {
      "bids": 1,
      "score": 1.0,
      "target": "0x62697264"
    },
    {
      "bids": 1,
      "score": 0.333,
      "target": "0x6b616e6761726f6f"
    },
    {
      "bids": 1,
      "score": 0.01,
      "target": "0x646f67"
    }
  ]
}

Links

Identifier: links, github:Userfeeds.algorithms.links

Time based algorithm that collects all claims with type link and sorts them according to formula:

period = 7 days

score = Sum(
  if now - claim.create_at > period:
    0
  else:
    claim.value / period * (now - claim.create_at)
)

Parameters

period

• string
• optional
• default: 7 days

Period of time decay to use during calculations

whitelist

• string
• optional

When present all claims type whitelist and claim.signature.creator equal to whitelist parameter will be used as filtering mechanism. Whitelisting claim needs to specify link claim claim.signature.signatureValue in it’s claim.target.

Whitelisting

Ads algorithm has special property that allows for arbitrary list filtering based on whitelist parameter passed to it. It works like that: Algorithm gathers all claims with type link intended for given context and compares their signature.signatureValue with claim.target value of claims with signature.creator equal to whitelist parameter. Example:

Link claims

context	type	target	signatureValue
ethereum:0xA	link	http://some.url/1	hashABC
ethereum:0xA	link	http://some.url/2	hashDEF
ethereum:0xA	link	http://some.url/3	hashGHI

Whitelist claims

target	creator
hashABC	ethereum:0xjohn
hashGHI	ethereum:0xjohn
hashABC	ethereum:0xalice
hashDEF	ethereum:0xbob

Above combination will produce two links with whitelist parameter equal to ethereum:0xjohn:

target
http://some.url/1
http://some.url/3

with whitelist equal to ethereum:0xalice it will produce one link:

target
http://some.url/1

and with ethereum:0xbob:

target
http://some.url/2

if you omit whitelist in api call the algorithm will produce full set of links:

target
http://some.url/1
http://some.url/2
http://some.url/3

Example data

{
  "items": [
    {
      "id": "0xdddf6852e4e239cba808ebdb620535a30189cb5e66b2a4bdbc46e41c844543d6",
      "score": 340000000000000000,
      "summary": "Has nice pipelines",
      "target": "https://bitbucket.org",
      "title": "Bitbucket"
    },
    {
      "id": "0xbc7dfcd2a2148cda1f763fb2d4713911c55c9f0d3fa56d62264c085f550a2586",
      "score": 321923548932000000,
      "summary": "Best social platform",
      "target": "https://github.com",
      "title": "GitHub"
    },
    {
      "id": "0x943e5c024f04acd0352095a09c71b4e5ce6e5d789da588ad6aacdec859917cd6",
      "score": 10000000000000000,
      "summary": "Userfeeds Platform",
      "target": "https://userfeeds.io/",
      "title": "Userfeeds"
    },
    {
      "id": "0xb4ed77fd554929f28c255279c307aedbd39510437e405693e5d84fd6a69b8fca",
      "score": 10000000000000000,
      "summary": "Status",
      "target": "https://status.im/",
      "title": "Status.im"
    }
  ]
}

Hold (TODO)

Identifier: hold, github:Userfeeds.algorithms.hold

Simple algorithm that collects all claims and sorts them according to cumulative hold. Hold is holdings*time for given token. If someone has 10 ETH for 1 day then we can say he has hold of value 10, if he still has 10 ETH during next 6 days his hold will raise to 70.

Writing ranking algorithms

The best way to start writing your own algorithms is to look at the built-in one at https://github.io/Userfeeds/algorithms

TODO: ^^ more effort here please...

Data available for ranking algorithms

Database dumps can be found here:

• Neo4j: http://some.s3.or.smemthing/neo4j/latest
• Postgresql: http://some.s3.or.something/postgresql/latest

Those data dumps are updated TODO:daily/hourly/constantly? and can be used for development and data analysis.

TODO: Import instructions

Note

Direct database access will be possible to our read-only instances for projects that require lower delays.

Claims

TODO

Balances and Aggregated data

TODO

Helper functions

TODO

Testing your algorithm

Importing database

Connecting your algorithm to Userfeeds Platform

TODO

Publishing your algorithm

Supported source are:

• Github
• Bitbucket
• Gist
• ...

TODO: ...

Registering algorithm on Userfeeds Platform

TODO