C / C++

12 October 2026 02:55

Then click the commit to the left of the line

C / C++

12 October 2026 01:23

https://github.com/unicode-org/icu/pull/3035

C / C++

12 October 2026 01:14

Ai says to do

git log -S "int32_t MAX_NESTED_LEVELS=0x03ff;" --reverse

C / C++

11 October 2026 22:48

The limits are fixed, the encoding/storage/packing is flexible

C / C++

11 October 2026 22:47

I sure will in the future 😄

C / C++

11 October 2026 22:41

Roman went off to reconsider all his base-10 numeric limits

C / C++

11 October 2026 22:39

If this relates to a recursive algorithm somewhere that actually has to recurse once per nesting level, it might be good to not permit 65535 stack frames

C / C++

11 October 2026 22:37

It just happens that there seems to be no use of extra fields within this value, so there's no need to shift or mask, since the 10 nestinglevel bits are the only bits in use

C / C++

11 October 2026 22:34

And you can have:
packed struct(u16) {
nestingLevel: u10,
reserved: u6,
}

C / C++

11 October 2026 22:33

If the limit is something round in decimal, you would obviously write it in decimal

C / C++

11 October 2026 22:29

It might as well be one where you make full utilization of the bits the value within that limit takes up

C / C++

11 October 2026 22:28

It could also mean that they just want to reserve some bits for future use

C / C++

11 October 2026 22:20

maybe it is a bitmask somewhere
but that somewhere is basically undefined.

C / C++

11 October 2026 22:19

yeah, that I do get, that makes sense

C / C++

11 October 2026 22:18

it's a simple depth limit.

Nobody explains why it's 1023 anywhere.
There's no bitwise operations involving it.
There's no place in the code that packs it into 10 bits

this is not a mask 🙂

C / C++

12 October 2026 02:54

Click "Blame" on github, since we already have the github link to the exact file

C / C++

12 October 2026 01:21

That's the worst way to find it

C / C++

12 October 2026 01:11

Maybe the commit that introduced it has an explanation？

C / C++

11 October 2026 22:48

One should not pass on future metadata slots

C / C++

11 October 2026 22:46

still a problem, the limits are just much higher 😄

C / C++

11 October 2026 22:39

(not a problem in Go ofc, since Go just allocates)

C / C++

11 October 2026 22:39

Come to think of it there might be another good reason to limit this value

C / C++

11 October 2026 22:36

But really you can view a normal u16 as the same thing

C / C++

11 October 2026 22:34

This is much simpler to think about in Zig ofc, where you just have u10

C / C++

11 October 2026 22:31

hmmm, now that I thought about it from that perspective it does make more sense now

C / C++

11 October 2026 22:28

Sometimes you just need to pick a limit

C / C++

11 October 2026 22:21

I mean, I get the part that it's 1023 so it fits exactly into 10 bits.
But there's zero justification and reasoning for it to fit in 10 bits, because it's always put in int16

C / C++

11 October 2026 22:20

but that limit constant declaration in hex makes zero sense to me 🙂 but alright, I don't think this discussion is worth continuing 😄

C / C++

11 October 2026 22:18

Maybe not in this repo, but elsewhere

C / C++

11 October 2026 22:18

0 -> 0000
3 -> 0011
f -> 1111
f -> 1111