Add uuids v6, v7, and v8 draft #55

mitch-kyle · 2024-07-10T23:53:01Z

Hello, I started working on this because I'm bored and have too much time on my hands. I just noticed you created a new branch for this feature. This is my WIP; I'm sure there are several mistakes but I wanted to share early to prevent duplicate effort. feel free to use any or none of it.

Add support for max, v6, v7, v8 uuids
Add helper functions for comparing the ordinality of multiple uuids

Draft RFC4122 ammendment
https://www.ietf.org/archive/id/draft-peabody-dispatch-new-uuid-format-04.html

- Add support for max, v6, v7, v8 uuids - Add helper functions for comparing the ordinality of multiple uuids Draft RFC4122 ammendment https://www.ietf.org/archive/id/draft-peabody-dispatch-new-uuid-format-04.html

mitch-kyle · 2024-07-10T23:58:25Z

src/clj_uuid.clj

           [java.util     UUID
-            Date]
-           [java.lang IllegalArgumentException]))
+            Date]))


I just ran clojure-sort-ns after adding a some namespaces out of habit

dco-lentz · 2024-07-12T14:39:21Z

very nice. and thanks for the contributions, i'm going through them now.

One concern I have is confusing or conflating these new features with the adopted standard, given the (abandoned?) draft status of this IETF memo. Especially insofar as it disagrees with or updates some details of the existing RFC-4122. Do you have any links to any current news about the standardization process?

mitch-kyle · 2024-07-12T15:09:14Z

Oh good call, I guess this is published RFC https://datatracker.ietf.org/doc/html/rfc9562

Not sure why that old draft is so ubiquitous and this one so hard to find.

Some notes from reading it:

RFC designation and numbering is off in the comments.
Section 6.2: There are 3 methods for the v7 monotonic counter. I've opted for the first but the others could be implemented as well if desired.

src/clj_uuid.clj

dco-lentz

really nice job. i left a few thoughts/questions, but this looks pretty sound.

I can go head and merge this progress into my 1.2.0 branch? or I can hold off if you wanted to keep going on this PR. I'd say my goals from here:

double check the monotonic+random spec vs. implementation
ensure backward compatibility (esp. protocol names etc)
maybe take a look at the naming /shadowing
supplement / improve internal code documentation
supplement / improve external code documentation
more tests? (you did a nice job on that though)

src/clj_uuid.clj

dco-lentz · 2024-07-16T14:15:14Z

README.md

@@ -631,6 +658,35 @@ _(function)_    `squuid []`
 >  time (seconds since 12:00am January 1, 1970 UTC) with the most
 >  significant 32 bits of the UUID

+_(function)_    `= [x]`


i'm normally a fan of concise syntax, but wondering if shadowing the clojure core numerical comparators is really the right thing to do here

dco-lentz · 2024-07-16T14:18:21Z

src/clj_uuid.clj

@@ -180,9 +193,9 @@



-(defprotocol UUIDRfc4122
+(defprotocol UUIDRfc9562


So, you think that RFC9562 should just replace/reimplement RFC4122 protcol? I wonder if it might be feasible to have this new protocol extend/supplement the other, in order to eliminate breaking changes for current users

The protocol is bound to it's old name in a var below; I was trying to avoid breaking changes while being precise with the name but if this breaks something then it should absolutely go back to the old name.

or something like that. I was actually thinking of breaking it into two protocols. Not that would make a difference in practice though. one difference (maybe) is if anyone has extended rfc4122UUID somewhere?

Creating two protocols is a fine option.

you're right, as is, implementers would have to define the new function in order to upgrade.

src/clj_uuid.clj

dco-lentz · 2024-07-16T14:29:08Z

src/clj_uuid.clj

@@ -396,16 +429,20 @@
    (URI/create (to-urn-string uuid)))

  (get-time-low ^long [uuid]
-    (bitmop/ldb #=(bitmop/mask 32 0)
-      (bit-shift-right (.getMostSignificantBits uuid) 32)))
+    (let [msb (.getMostSignificantBits uuid)]


since these guys are special casing types under the hood, i wonder if it would help readability if I/we added some docstrings to these

documenting the differences in resolution between v1 and v6 wouldn't be a bad thing

yeah, i'm going to spend some time on some of this i always wanted this library to have good documentation, since UUIDs seem so poorly understood

dco-lentz · 2024-07-16T14:43:07Z

src/clj_uuid.clj

+
+(defn max
+  "Generates the maximum UUID, ffffffff-ffff-ffff-ffff-ffffffffffff."
+  ^java.util.UUID


i know this is consistent withnull and the name of the thing is "max uuid". unfortunately as yet i don't have a better name.

It matches the documentation and is consistent with the null function. But I am open to giving this a different name because we could conceivably add min and max functions for uuids. Maybe maximum?

i dunno. we're sort of painted into a corner with the nomenclature. in common lisp maybe it would be analogous to t -- true, the top type, opposite of nil? one idea anyway. maybe a bad one

I think the name should follow from the RFC

ya fair point

dco-lentz · 2024-07-16T14:48:11Z

src/clj_uuid.clj

+  guaranteed to be unique and thread-safe regardless of clock
+  precision or degree of concurrency.  Creation of v6 UUID's does not
+  require any call to a cryptographic generator and can be
+  accomplished much more efficiently than v3, v4, v5, or squuid's.  A


wouldn't this then also be more efficient than v7?

yes, v7 could be added to this list. v7 is still preferred by the RFC because it's less guessable.

i see, gotcha. that's good context I'll add, as well. I really want to also do some micro benchmarking and timing

dco-lentz · 2024-07-16T14:57:47Z

src/clj_uuid.clj

+      (f y (first more)))
+    false))
+
+(defn =


one idea regarding the >, =, etc shadowing. -- what if instead we added variable arity under the hood to uuid= and uuid> ?

the issue is as far as I know, protocol functions can't have variadic arguments.
https://clojure.atlassian.net/browse/CLJ-888

shadowing core functions is matter of taste, I like it because clojure.core is a huge library and it takes all the good names :P. We see official clojure libraries like core.async override many core names.

Here's a video of Rich talking about how he loves namespaces because he got tired of coming up with new names: https://youtu.be/oyLBGkS5ICk?feature=shared&t=662

oh, right, variadic protocol arguments. I see your point. Well, let me see if i can come up with alternative. otherwise, not the end of the world unless anyone :use s the namespace or :refer :all

:refer :all is considered an anti-pattern for this reason. This is why I like clojure's use of namespaces over scheme's default use-module semantics

dco-lentz · 2024-07-16T15:20:05Z

test/clj_uuid/clock_test.clj

-
-
-
+(deftest check-monotonic-unix-time-and-random-counter


yeah the function is kind of a mouthful but at least it's specific :P I'm open to alternatives

mitch-kyle · 2024-07-16T16:25:48Z

src/clj_uuid.clj

@@ -324,9 +352,11 @@
    "Return the unique URN URI associated with this UUID."))


+;; For backwards compatibility
+(def UUIDRfc4122 UUIDRfc9562)


This should make the protocol name change backwards compatible but if I'm mistaken we should put the old protocol name back.

mitch-kyle · 2024-07-16T16:38:28Z

I can go head and merge this progress into my 1.2.0 branch?

Absolutely, I don't think I have anything else to add to this PR.

I'll still be available to make or review changes if you need it.

dco-lentz · 2024-07-16T19:14:23Z

can we connect over email? i'd love to know more about what you're up to

mitch-kyle · 2024-07-16T19:27:25Z

Sure, my email is mitch.tux@gmail.com

Add uuids v6, v7, and v8 draft

b0fc196

- Add support for max, v6, v7, v8 uuids - Add helper functions for comparing the ordinality of multiple uuids Draft RFC4122 ammendment https://www.ietf.org/archive/id/draft-peabody-dispatch-new-uuid-format-04.html

mitch-kyle commented Jul 10, 2024

View reviewed changes

mitch-kyle force-pushed the 53-v7-support branch from 5264d83 to 43df42b Compare July 11, 2024 00:01

mitch-kyle added 2 commits July 10, 2024 21:01

Clean up

43df42b

Fix resolution of v7 counter

5483227

mitch-kyle commented Jul 12, 2024

View reviewed changes

src/clj_uuid.clj Outdated Show resolved Hide resolved

mitch-kyle added 2 commits July 12, 2024 15:51

Update documentation from Rfc4122 to Rfc9562

a9c0fcc

Fix v6 get-timestamp

28fd247

mitch-kyle force-pushed the 53-v7-support branch from 7a2c777 to 855ec30 Compare July 12, 2024 22:40

mitch-kyle added 2 commits July 12, 2024 19:41

Add get-unix-time for accessing v7 timestamp

855ec30

clarify some docstrings

62b962e

mitch-kyle commented Jul 13, 2024

View reviewed changes

src/clj_uuid.clj Show resolved Hide resolved

dco-lentz approved these changes Jul 16, 2024

View reviewed changes

mitch-kyle commented Jul 16, 2024

View reviewed changes

Update v6 docstring

4b973e5

dco-lentz merged commit 8e57ecc into danlentz:53-v7-support Jul 16, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add uuids v6, v7, and v8 draft #55

Add uuids v6, v7, and v8 draft #55

mitch-kyle commented Jul 10, 2024

mitch-kyle Jul 10, 2024 •

edited

Loading

dco-lentz commented Jul 12, 2024

mitch-kyle commented Jul 12, 2024 •

edited

Loading

dco-lentz left a comment

dco-lentz Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

dco-lentz Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024 •

edited

Loading

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024 •

edited

Loading

dco-lentz Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

dco-lentz Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

dco-lentz Jul 16, 2024

mitch-kyle Jul 16, 2024

mitch-kyle Jul 16, 2024

mitch-kyle commented Jul 16, 2024

dco-lentz commented Jul 16, 2024

mitch-kyle commented Jul 16, 2024

		@@ -180,9 +193,9 @@



		(defprotocol UUIDRfc4122
		(defprotocol UUIDRfc9562




		(deftest check-monotonic-unix-time-and-random-counter

Add uuids v6, v7, and v8 draft #55

Add uuids v6, v7, and v8 draft #55

Conversation

mitch-kyle commented Jul 10, 2024

mitch-kyle Jul 10, 2024 • edited Loading

Choose a reason for hiding this comment

dco-lentz commented Jul 12, 2024

mitch-kyle commented Jul 12, 2024 • edited Loading

dco-lentz left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mitch-kyle Jul 16, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mitch-kyle Jul 16, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mitch-kyle commented Jul 16, 2024

dco-lentz commented Jul 16, 2024

mitch-kyle commented Jul 16, 2024

mitch-kyle Jul 10, 2024 •

edited

Loading

mitch-kyle commented Jul 12, 2024 •

edited

Loading

mitch-kyle Jul 16, 2024 •

edited

Loading

mitch-kyle Jul 16, 2024 •

edited

Loading