feat: add a text output encoding for the stats provide command

[aschmahmann]

This adds a text output encoding to the ipfs stats provide command for easier glance value

It looks something like:

❯ ipfs stats provide
TotalProvides: 7084
AvgProvideDuration: 15.358954ms
LastReprovideDuration: 58.3776802s
LastReprovideBatchSize: 3542

It could be a little prettier, but here's a first stab so at least we have something.

[lidel]

I was trying to see how it behaves irl and numbers were stuck at zero:

TotalProvides: 0
AvgProvideDuration: 0s
LastReprovideDuration: 0s
LastReprovideBatchSize: 0

I thought that might be because the repo I used had English Wikipedia so a lot of blocks, so I left it for a night and same numbers in the morning. Looks like stats are not updated for some reason?

Re-tested with new repo (ipfs init + all ports set to 0 to avoid conflict with my other node) and found a panic after 1 minute:

$ ipfs daemon
....
Daemon is ready
panic: runtime error: integer divide by zero

goroutine 4522104 [running]:
github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend.func1(0xc007bb14b0, 0xc0119c1ca0, 0x2, 0x4, 0xc007bb14a0, 0x0, 0xc0119c1b80, 0x16, 0x16, 0xc01390d7d0, ...)
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:981 +0x451
created by github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:977 +0x256
ipfs daemon  347.53s user 59.88s system 112% cpu 6:00.68 total

I was able to reproduce the panic after restarting the daemon, failed again, exactly after 1 minute, with:

panic: runtime error: integer divide by zero

goroutine 4450306 [running]:
github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend.func1(0xc005f86240, 0xc003f68800, 0x2, 0x4, 0xc005f86220, 0x0, 0xc003f686e0, 0x16, 0x16, 0xc014aab380, ...)
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:981 +0x451
created by github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:977 +0x256
panic: runtime error: integer divide by zero

goroutine 4450250 [running]:
github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend.func1(0xc005f86240, 0xc003f68700, 0x2, 0x14, 0xc005f86220, 0x0, 0xc003f686e0, 0x16, 0x16, 0xc014aab380, ...)
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:981 +0x451
created by github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:977 +0x256
panic: runtime error: integer divide by zero

goroutine 4450253 [running]:
github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend.func1(0xc005f86240, 0xc003f68760, 0x2, 0xe, 0xc005f86220, 0x0, 0xc003f686e0, 0x16, 0x16, 0xc014aab380, ...)
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:981 +0x451
created by github.com/libp2p/go-libp2p-kad-dht/fullrt.(*FullRT).bulkMessageSend
	pkg/mod/github.com/libp2p/[email protected]/fullrt/dht.go:977 +0x256
ipfs daemon  556.69s user 60.15s system 171% cpu 6:00.65 total

[aschmahmann]

@lidel the panic you encountered is unrelated to this and should be resolved by libp2p/go-libp2p-kad-dht#719. I'll rebase this PR on #8156 to make this one more testable though

[aschmahmann]

@lidel can you think of a reason why we should/shouldn't have some -H or --human flag here to enable the humanized numbers?

I see we do it in other commands (ipfs repo stat, ipfs bitswap stat), although I'm not sure why we wouldn't want it to always be human if --enc=Text and more machine friendly with --enc=json.

[lidel]

Good question. I think it is a matter of things growing organically and not having a clear convention from the beginning.
I feel its most useful to optimize text for quick eyeballing, and show unmodified values in JSON, but its just personal preference.
Happy to add --human flag if you feel it's more in line with other commands.

Anyway, I rounded Durations to miliseconds, it looks like this:

TotalProvides:          73  (73)
AvgProvideDuration:     223ms
LastReprovideDuration:  16.32s
LastReprovideBatchSize: 73  (73)

Perhaps we should hide "(val)" if it is the same as SI (values under 1k)

[aschmahmann]

Happy to add --human flag if you feel it's more in line with other commands.

Ya, I guess we might as well for sanity's sake. Although I'd be in favor of pushing us to a place where Text is human readable + flexible and other formats are more machine friendly.

Anyway, I rounded Durations to miliseconds, it looks like this:

I wouldn't quite do that, as the number of provides get larger the average duration shrinks and can drop into the single digit ms range. e.g. I recently got this log output finished providing of 1060189 keys. It took 26m46.469890381s with an average of 1.515267ms per provide. Maybe it's easy to just use a couple of decimal places instead of anything fancier like significant digits.

[lidel]

Given this is experimental opt-in, should be good enough.
We can refine output when/if we decide to move this outside of Experiments.

I pushed small refactor for text representation (12ebbe5):

• hides "(n)" when value is same as SI,
• removed space separator from SI notation ("3k" instead of "3 k")

TotalProvides:          35k (35,171)
AvgProvideDuration:     2.728ms
LastReprovideDuration:  1m35.952581s
LastReprovideBatchSize: 35k (35,171)