zero padding display hint diverges from core::fmt

[Urhengulas]

This code:

defmt::info!("{:00x}", 42);
defmt::info!("{:01x}", 42);
defmt::info!("{:02x}", 42);
defmt::info!("{:03x}", 42);
defmt::info!("{:04x}", 42);
defmt::info!("{:05x}", 42);
defmt::info!("{:06x}", 42);

results in these logs:

0 INFO  0x2a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:8
1 INFO  0x2a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:9
2 INFO  0x2a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:10
3 INFO  0x2a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:11
4 INFO  0x2a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:12
5 INFO  0x02a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:13
6 INFO  0x002a
└─ hello::__cortex_m_rt_main @ src/bin/hello.rs:14

---

I'd expect that a display hint :04x results in 0x002a for 42, but it currently seems to include the 0x in the length and only results in 0x2a. 06x will currently result in the desired 0x002a, but I would expect 0x000002a.

[Urhengulas]

@derekdreery Would you like to look into this?

[jonas-schievink]

This matches the behavior of core::fmt, so it might be better to keep it

[Urhengulas]

Really? That is weird, but okay with me if it is consistent 🤔

[jonas-schievink]

Hmm, actually our output for {:06x} matches {:#06x} with core::fmt - it doesn't include the 0x unless you use #, so maybe it's fine if we change how the padding works too?

[Urhengulas]

Hmm, actually our output for {:06x} matches {:#06x} with core::fmt - it doesn't include the 0x unless you use #, so maybe it's fine if we change how the padding works too?

That is also what I just tested and was about to write 😅

This code

println!("{:02X}", 42);
println!("{:#02X}", 42);
println!("{:03X}", 42);
println!("{:#03X}", 42);
println!("{:04X}", 42);
println!("{:#04X}", 42);

results in

2A
0x2A
02A
0x2A
002A
0x2A

I think consistency is helpful to help people not getting confused, about display hints. I'd rather vote for matching the behavior, and explude the 0x.

[richard-uk1]

When I made this, I tried to merge the beahvior of core::fmt and the existing defmt formatting. Happy to do a PR to change it if that's the consensus.

…

On Sun, May 23, 2021 at 10:38 PM Johann Hemmann ***@***.***> wrote: Hmm, actually our output for {:06x} matches {:#06x} with core::fmt - it doesn't include the 0x unless you use #, so maybe it's fine if we change how the padding works too? That is also what I just tested and was about to write 😅 I think consistency is helpful to help people not getting confused, about display hints. I'd rather vote for matching the behavior, and explude the 0x. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#491 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAKT4XVCNMM5MKRHHALTTETTPFYUDANCNFSM45MDCPIQ> .

[Urhengulas]

@derekdreery: I would be happy if you go forward to adapt the output to match core::fmt! 😄

[richard-uk1]

Hi, so there are a few differences between defmt and core::fmt:

1. In core::fmt you have :x and :X which will use lower and upper case a-f respectively.
2. In core::fmt, the alternative :#x rather than :x includes the 0x at the beginning, in defmt 0x is always included.

The might be others. Would you like me to change behaviour to match for 1 and 2, just 1, or something else?

[Urhengulas]

Number one is already the case for defmt, isn't it? Thinking about it again I think including the 0x isn't that bad of an idea, therefore I think no action is required.

[richard-uk1]

@Urhengulas are you sure that having :x and :X differ to core::fmt is good? I've currently got a MAC address I'd like to format as AA:BB:11:22:33:44 but currently can't do this because I get 0xAA:0xBB:0x11:0x22:0x33:0x44. In core::fmt you can still get the 0x prefix by using the alternative formatter (:#x and :#X).

People will be familiar with how core::fmt works and making defmt work the same will reduce cognitive overhead.

[Urhengulas]

I see. I mainly thought about logging single hex numbers, but for something like a mac-address it makes sense to be able to exclude the 0x. I then would agrre, that completely matching the core::fmt behavior makes sense 😄