Data versioning done right
A data versioning library.
$ npm install lineage
The most basic version data type you can use is a Number or Date.
var lineage = require'lineage'protocol = lineageprotocol;// Require this to extend Number with the version protocol provided by lineage.require'lineage/lib/types/number';var oldVersion = 1;var newVersion = protocolincroldVersion; // => 2var comparison = protocolcompareoldVersion newVersion;console.logcomparison === protocolconstsLT; // truecomparison = protocolcompareoldVersion oldVersion;console.logcomparison === protocolconstsEQ; // truecomparison = protocolcomparenewVersion oldVersion;console.logcomparison === protocolconstsGT; // true
Dates act similarly to numbers except version incrementing a Date just means returning the Date right now.
var lineage = require'lineage'protocol = lineageprotocol;// Require this to extend Date with the version protocol provided by lineagerequire'lineage/lib/types/number';var oldVersion = ;setTimeoutvar newVersion = protocolincroldVersion; // => <Date>var comparison = protocolcompareoldVersion newVersion;console.logcomparison ==== protocolconstsLT; // truecomparison = protocolcompareoldVersion oldVersion;console.logcomparison protocolconstsEQ; // truecomparison = protocolcomparenewVersion oldVersion;console.logcomparison === protocolconstsGT; // true1;
For distributed systems, the ideal data structure to use is a vector clock, also known as a Lamport Clock.
Lineage provides vector clocks out of the box.
Why vector clocks are critical when you want to version data in a distributed system is more involved, and this README will point the reader instead to the canonical paper on Lamport clocks.
What a vector clock does behind the scenes is more straightforward. Vector clocks extend the concept of the Number as a counter that represents the version. Vector clocks are a set of counters, where each counter is associated with the actor/client/node that updated the vector clock. Every time a particular actor updates a vector clock, the counter it is associated with is incremented.
var lineage = require('lineage'), protocol = lineage.protocol, Clock = require('lineage/lib/types/clock');var oldVersion = new Clock(); // => <Clock>// Compared to Number as a version, Clocks as versions must associate every// update to a version with an actor id. Here that actor id is 'actor-A'var newVersion = protocol.incr(oldVersion, 'actor-A');// Vectors updated by an actor compare to other vectors just how you would// expect them to.var comparison = protocol.compare(oldVersion, newVersion);console.log(comparison === protocol.consts.LT); // truecomparison = protocol.compare(oldVersion, oldVersion);console.log(comparison === protocol.consts.EQ); // truecomparison = protocol.compare(newVersion, oldVersion);console.log(comparison === protocol.consts.GT); // true
Up until this point, it would appear as though Clock versions compare to each other in the same way that Number versions compare to each other. So why even have a different data structure to use for versions?
In a contrived real world scenario, consider a person in San Francisco who owns a calendar. If she adds an event to the calendar each day, then the calendar on any given day is a later version than the calendar the day before. In this scenario, using a Number as a counter version works perfectly well.
Things become more interesting when we involve another actor updating the version -- in this case, we have a distributed system.
Suppose the San Francisco citizen makes a copy of her calendar and sends it to a man in China. Now suppose that on the first day that they both have identical calendars, each of them decide to update their calendars. How would you compare the 2 calendars? They are older than their respective versions, but how do they compare to each other? They are not at equivalent versions, but one does not also have a greater or less version than the other. In this case, we say the two versions are concurrent.
var lineage = require'lineage'protocol = lineageprotocolClock = require'lineage/lib/types/clock';var sfVersion = ; // => <Clock>var chineseVersion = ; // => <Clock>protocolincrsfVersion 'sf-person';protocolincrchineseVersion 'chinese-person';var comparison = protocolcomparesfVersion chineseVersion;console.logcomparison === protocolconstsCONCURRENT; // true
In distributed systems, we can reconcile data at concurrent versions by resolving the given data conflict inherent in the concurrent versions of data and then merge the two versions into a new version that is greater than the two versions.
In the contrived real world example, this could occur if the Chinese man sends the San Francisco woman his calendar. She could then make a 3rd calendar that included both her and his events. This calendar would then be considered at a greater version than the two prior calendars because it reflects a calendar that takes into account the historic set of changes to both calendars.
Continuing with our code from the last example:
var reconciledVersion = protocolmergesfVersion chineseVersion;comparison = protocolcomparereconciledVersion sfVersion;console.logcomparison === protocolconstsGT; // truecomparison = protocolcomparereconciledVersion chineseVersion;console.logcomparison === protocolconstsGT; // true
Vector clocks are essential to handle data versioning in a distributed system. However, they can also grow in an unbounded way if the number of actors who can update the version grows in an unbounded way. If there are 100 actors, the clock will consist of 100 counters behind the scenes.
In practice, what is often used instead is a variation known as an annealed vector clock, that limits the number of actors to a maximum. If a vector clock has already been updated by a maximum number of actors, then a new actor who decides to update that version will effectively replace the counter that has not been updated the longest.
var lineage = require'lineage'protocol = lineageprotocolLengthAnnealedClock = require'lineage/lib/types/length_annealed_clock';var versionA = 2; // => <LengthAnnealedClock>var versionB = 2; // => <LengthAnnealedClock>protocolincrversionA 'actor-A';protocolincrversionA 'actor-B';protocolincrversionB 'actor-A';protocolincrversionB 'actor-B';var comparison = protocolcompareversionA versionB;console.logcomparison === protocolconstsEQ;// Here we update the version with a new actor, but the version clock is// already at its maximum number of counters.protocolincrversionB 'actor-C';comparison = protocolcompareversionA versionB;// Normally, in a non-annealed Clock, this comparison would place versionA as// LT (less than) versionB.// When we anneal a Clock, we can lose information of what other actors have// done to a version historically. Because of that, versionB actually counts as// being CONCURRENT to versionA.console.logcomparison === protocolconstsCONCURRENT;
Copyright (c) 2012 by Brian Noguchi and Nate Smith
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.