Feature request: Calculate statistics on a (wrapped) stream

[vshymanskyy]

Could be interesting to be able to grab some stats of the stream usage:

• Total TX/RX bytes count
• Total TX/RX packets count (writes)
• Min/Max/Average read/write size

The only difficulty I see here: Some functions (i.e. readBytes) could be implemented via read() on some streams, but it can also be implemented directly by other implementations. Not sure if there's a clean way to handle it.

[bblanchon]

Hi @vshymanskyy,

Thank you for this suggestion; I think this is a good idea.

We could name the new classes:

• CountingPrint
• CountingStream
• WriteCountingStream
• ReadCountingStream
• etc.

The members could be:

• readBytes()
• readCount()
• minReadSize()
• maxReadSize()
• avgReadSize()
• writtenBytes()
• writeCount()
• minWriteSize()
• maxWriteSize()
• avgWriteSize()

How does that sound?

Best regards,
Benoit

[vshymanskyy]

Sounds cool. But some names may be ambiguous esp. readBytes(). Maybe prefix all of them with total, min, max, avg ?

[bblanchon]

You're right!
Maybe something like this, then:

read	write
readCount()	writeCount()
avgReadSize()	avgWriteSize()
maxReadSize()	maxWriteSize()
minReadSize()	minWriteSize()
totalReadSize()	totalWriteSize()

[bblanchon]

I'm not sure how to handle error cases: should we count them, or ignore them?

For example, if upstream's write() function returns 0, should I increment writeCount? should I set minWriteSize to 0?

[dpharris]

Maybe readErrors() and writeErrors()?

[mhaberler]

for such jobs I like the header-only stats classes by https://www.johndcook.com/blog/skewness_kurtosis/

he has several such classes on his website, real collectors items - I collect them here plus my adaptations: https://github.com/mhaberler/Stats