As described in the ulid specification repo, and slightly edited here, UUID use can be suboptimal for many uses-cases because:
- It is not the most character efficient way of encoding 128 bits of randomness
- UUID v1/v2 is impractical in many environments, as it requires access to a unique, stable MAC address
- UUID v3/v5 requires a unique seed and produces randomly distributed IDs, which can cause fragmentation in many data structures
- UUID v4 provides no other information than randomness which can cause fragmentation in many data structures
Instead, an alternative is proposed in ULID:
ulid() // 01ARZ3NDEKTSV4RRFFQ69G5FAVwith the following properties:
- 128-bit compatibility with UUID
- 1.21e+24 unique ULIDs per millisecond
- Lexicographically sortable!
- Canonically encoded as a 26 character string, as opposed to the 36 character UUID
- Uses base32 by Crockford for better efficiency and readability (5 bits per character)
- Case insensitive
- No special characters (URL safe)
- Monotonic sort order (correctly detects and handles the same millisecond)
01AN4Z07BY 79KA1307SR9X4MV3
|----------| |----------------|
Timestamp Randomness
48bits 80bits
Timestamp
- 48 bit integer
- UNIX-time in milliseconds
- Will not run out of space until the year 10889 AD.
Randomness
- 80 bits
- Cryptographically secure source of randomness, if possible
The left-most character must be sorted first, and the right-most character sorted last (lexical order). The default ASCII character set must be used. Within the same millisecond, sort order is not guaranteed.
The following functions are implemented:
ts_generate: Generate ULIDs from timestampsULIDgenerate: Generate ULIDsunmarshal: Unmarshal a ULID into a data frame with timestamp and random bitstring columnsulid: Alias forULIDgenerate
The package can be installed from CRAN via install.packages("ulid").
Development versions can also be installed from this repository or from
r-universe via
r <- c('https://eddelbuettel.r-universe.dev', 'https://cloud.r-project.org')
install.packages('ulid', repos = r)ulid::ULIDgenerate()
## [1] "0001EKRGEEV98QP062VNRX31P2"
(u <- ulid::ULIDgenerate(20))
## [1] "0001EKRGEEV5XMP54RRRWAK318" "0001EKRGEEKX7VC0PF75AZJXHP"
## [3] "0001EKRGEEXENNCQEH4KCH8QAD" "0001EKRGEEY41HJ6GMXRV1BQBA"
## [5] "0001EKRGEE6HVD7ACWZ52MTVCJ" "0001EKRGEEQWXMPXGC0DGQN32B"
## [7] "0001EKRGEE6W13BK92EF1RXYT7" "0001EKRGEE5A31H38NJFGTK8PC"
## [9] "0001EKRGEEG2GXS53QY9F3M0A9" "0001EKRGEEDA3Y6Y0T52WTS6RM"
## [11] "0001EKRGEE5WS2S3D9KY3F5H9Y" "0001EKRGEE24SZW5NATAADAY9Q"
## [13] "0001EKRGEEBEG51QCKXPM8ZS16" "0001EKRGEE1ZC1QY7RCJR9VJ0B"
## [15] "0001EKRGEECJ50Z4FXM4HW6XWG" "0001EKRGEEER84JP8WTXV5DWV8"
## [17] "0001EKRGEEW3ABA82GZSRXN1RB" "0001EKRGEEAA60CYFGR8832JD6"
## [19] "0001EKRGEE6W5ARCFHH6T75FPZ" "0001EKRGEE5WT4XNP7NS69BM3X"
unmarshal(u)
## ts rnd
## 1 2019-07-27 08:21:34 V5XMP54RRRWAK318
## 2 2019-07-27 08:21:34 KX7VC0PF75AZJXHP
## 3 2019-07-27 08:21:34 XENNCQEH4KCH8QAD
## 4 2019-07-27 08:21:34 Y41HJ6GMXRV1BQBA
## 5 2019-07-27 08:21:34 6HVD7ACWZ52MTVCJ
## 6 2019-07-27 08:21:34 QWXMPXGC0DGQN32B
## 7 2019-07-27 08:21:34 6W13BK92EF1RXYT7
## 8 2019-07-27 08:21:34 5A31H38NJFGTK8PC
## 9 2019-07-27 08:21:34 G2GXS53QY9F3M0A9
## 10 2019-07-27 08:21:34 DA3Y6Y0T52WTS6RM
## 11 2019-07-27 08:21:34 5WS2S3D9KY3F5H9Y
## 12 2019-07-27 08:21:34 24SZW5NATAADAY9Q
## 13 2019-07-27 08:21:34 BEG51QCKXPM8ZS16
## 14 2019-07-27 08:21:34 1ZC1QY7RCJR9VJ0B
## 15 2019-07-27 08:21:34 CJ50Z4FXM4HW6XWG
## 16 2019-07-27 08:21:34 ER84JP8WTXV5DWV8
## 17 2019-07-27 08:21:34 W3ABA82GZSRXN1RB
## 18 2019-07-27 08:21:34 AA60CYFGR8832JD6
## 19 2019-07-27 08:21:34 6W5ARCFHH6T75FPZ
## 20 2019-07-27 08:21:34 5WT4XNP7NS69BM3X
(ut <- ts_generate(as.POSIXct("2017-11-01 15:00:00", origin="1970-01-01")))
## [1] "0001CZM6DG2THKSAX3F1SF30E7"
unmarshal(ut)
## ts rnd
## 1 2017-11-01 15:00:00 2THKSAX3F1SF30E7As per issue #13 on the upstream repo, time is actually
encoded mostly as time_t leading to second rather than millisecond resolution. Two patches by
Chris Brove also collected in his fork improve on this by using
std::chrono objects internally. In release 0.4.0, we have switches to his fork and extended the
wrapper functions to support this:
> library(ulid)
> gen_ulid <- \(sleep) replicate(5, {Sys.sleep(sleep); generate()})
> u <- gen_ulid(.1)
> df <- unmarshal(u)
> data.table::data.table(df)
ts rnd
<POSc> <char>
1: 2024-05-30 16:38:28.588 CSQAJBPNX75R0G5A
2: 2024-05-30 16:38:28.688 XZX0TREDHD6PC1YR
3: 2024-05-30 16:38:28.789 0YK9GKZVTED27QMK
4: 2024-05-30 16:38:28.890 SC3M3G6KGPH7S50S
5: 2024-05-30 16:38:28.990 TSKCBWJ3TEKCPBY0
>Suyash Verma wrote the C++ header library ulid.
Chris Bove updated internals to permit sub-second resolution in his fork.
Bob Rudis created the R package, prepared versions 0.1.0 and 0.2.0, and released version 0.3.0 to CRAN.
Dirk Eddelbuettel has been maintainer since release 0.3.1.
The package is licensed under the MIT License