const Bronze = require('bronze')
const idGen = new Bronze({name: 'example'})
const id = idGen.generate()
// 1482810226160-0-14210-example-1a
$ bronze --name example
# 1496196550327-31-7887-example-1a
$ npm install bronze
CLI only:
$ [sudo] npm install bronze -g
- designed for distributed and singleton systems
- no time-based (
UUID1
) or random (UUID4
) collisions - collision resistant - safely generate up to 9,007,199,254,740,991 ids within a single millisecond
- fast - can generate an id in less than .05ms - even on old hardware
- can be conveniently sorted by
timestamp
orname
- can be used as a module or via CLI
- supports Node 6+ and browsers
const Bronze = require('bronze')
const idGen = new Bronze()
idGen.generate()
// > 1483113483923-0-12179-localhost-1a
idGen.generate()
// > 1483113483923-1-12179-localhost-1a
// continue to create ids throughout your code
A spec determines what goes into an id and how its information is sorted.
Type 1
- contains a timestamp
in milliseconds, a sequence
number, a process id (PID
), and a name
1a
is ordered bytimestamp
-sequence
-pid
-name
- useful for timestamp-based sorting
- ex:
1482981306438-0-5132-example-1a
1b
is ordered byname
-pid
-timestamp
-sequence
- useful for name-based sorting
- ex:
example-5132-1482981317498-0-1b
Bronze
CLASS
-
static get
defaultName
STRING- the default name for new instances (if not provided)
-
static get
defaultSpec
STRING- the default spec for new instances
-
static get
specs
OBJECT- the default name
-
static
parse
(id)-
parses an id to JSON-compatible object
-
id
STRING required- an id to parse
-
-
new Bronze (options)
- Example:
const idGen = new Bronze({name: 'example', sequence: 1})
-
options
OBJECTsequence
INTEGER- the number of ids generated.
- convenient for continuing creating ids where you left off (no pre-increment required)
- if invalid falls back to default
- default = 0
pid
INTEGER- a process id to number to use for generated ids
- if invalid falls back to default
- default = process.pid, else 0
name
STRING- a unique name for the generator - replaces slashes (\/) with underscores (_)
- IMPORTANT - do not use the same name at the same time on different machines
- if invalid falls back to default
- default = process.env.HOSTNAME, else constructor.defaultName
spec
STRING- set the spec
- if invalid falls back to default
- default = constructor.defaultSpec
-
instance OBJECT
-
sequence
INTEGER- the current sequence
-
pid
INTEGER- the pid in use
-
name
STRING- the parsed name to be used in ids
-
nameRaw
STRING- the raw, pre-parsed name
-
spec
STRING- the spec in use
-
generate
([options])- generates a new id
- Example:
const idGen = new Bronze() const id = idGen.generate() // > 1482810226160-0-14210-localhost-1a
-
options
OBJECT optionaljson
BOOLEAN- returns results as JSON-compatible object instead of STRING
- default = false
-
return id STRING | OBJECT
- The generated id
- returns an object if
options.json === true
-
nextSequence
()- manually advances the sequence
- Example:
const idGen = new Bronze() idGen.nextSequence()
-
Using bronze in the browser is pretty straight-forward. As of webpack 3, no special loaders are required to use bronze. Since most browser environments do not support the process
object (with the exception of Electron, NW.js, and the like), you should pass the pid
and name
options to the constructor, like so:
new Bronze({pid: 1, name: 'browser'})
If you are using bronze in a distributed environment you should verify the generated name
via Bronze.parse
in a trusted space, such as the server-side.
The CLI uses the module under the hood.
Usage: bronze [options]
Options:
--sequence INT Set the counter for the number of ids generated
By default will use sequence file (sequence path)
If set sequence file will not be updated
--pid INT Set the process id for generated ids
--name STRING A unique name for the generator
Any slashes will be replaced with underscores
--spec STRING Set the spec
--gen, -g INT The number of ids to create. Must be >= 0.
Default = 1
--list-specs Get the specs available from this version of bronze
--sequence-dir STRING Set the sequence directory
Will attempt to create if not exist
--sequence-dir-reset Sets the sequence back to 0
File isn't created if it doesn't exist
--help, -h
--version, -v
While developing a distributed system using UUID1
and UUID4
we found that we would run into collisions as we scaled.
UUID1
(timeuuids) can have collisions within 100ns, which can happen when migrating/importing data in bulkUUID4
can have collisions at random. While the risk is reasonably small, a chance of data loss does not sit well with us.
- no collisions
- remove 'randomness' as an element for unique identifiers, which have the possibility of collisions
- predictability > random/entropy
- even a slim chance is unacceptable, mission-critical apps
- keep it simple
- fast - should be able to uniquely identify data in real-time apps
- having different sorting/grouping options would be nice
- a built-in counter (sequence) would be nice for statistics, especially if you can restore it after reboot
- compatible with Node's
cluster
module with no additional setup - eliminate need for performance-draining read-before-writes to check if a unique identifier exists
- reduced need for counters and auto-incrementing rows
- Node 6+ is required for 1.x+.
name
may contain any character, including dashes, but slashes (\/) will be replaced with underscores (_).- This allows the opportunity for an id used as a cross-platform filename with little concern.
- Every machine in your distributed system should use a unique
name
. If every machine has a unique hostname (process.env.HOSTNAME
) you should be fine. - To avoid collisions:
- Do not reuse a
name
on a different machine within your distributed system's range of clock skew. - Be mindful when changing a system's clock - if moving the clock back temporarily change the
name
- Only one instance of Bronze should be created on a single process (
PID
) to avoid collisions.- Each worker spawned from Node's
cluster
module receives its ownPID
so no worries here.
- Each worker spawned from Node's
- Do not reuse a
- Sequence counter automatically resets after reaching
Number.MAX_SAFE_INTEGER
(9007199254740991) - Without string parsing, timestamp sorting (
spec 1a
) is supported up to2286-11-20T17:46:39.999Z
(9999999999999) - Using with databases, such as Cassandra:
- most
text
data types should do the trick
- most
-
CLI
- add
--parse
option- JSON output
- add
-
Nested IDs
const id1 = idGen.generate({name: 'example'}) console.log(id1) // > 1482810226160-0-14210-example-1a // Nested idGen.name = id1 const id2 = idGen.generate() console.log(id2) // > 1482810226222-1-14210-1482810226160-0-14210-example-1a-1a Bronze.parse('1482810226222-1-14210-1482810226160-0-14210-example-1a-1a') // { valid: true, // timestamp: 1482810226222, // sequence: 1, // pid: 14210, // name: '1482810226160-0-14210-example-1a', // spec: '1a' } // can also nest id2, id3, id4, id5, ...idN
- the issue the moment is the possibility of collisions (no unique name)
- would be pretty cool - imagine the convenience of nesting a video_id into a comment_id
-
See more in FUTURE.md