Re: [go-nuts] log package implementation question

[hanisch]

Thanks Ian!  I see that at some point thereafter Logger's buf member must 
have been changed from a bytes.Buffer (using WriteString, WriteByte, etc.) 
to a []byte (using append) as well.

Thanks again!


On Wednesday, June 7, 2017 at 3:32:52 PM UTC-4, Ian Lance Taylor wrote:
>
> On Wed, Jun 7, 2017 at 11:12 AM,  > wrote: 
> > 
> > I was perusing the code for the standard log package (src/log/log.go - 
> The 
> > Go Programming Language) and was wondering why the formatHeader method 
> (line 
> > 92) takes a pointer to byte slice since the only call to this method 
> line 
> > (163) calls it with the address of its own byte slice member (and it 
> only 
> > has one such member). 
> > 
> > To be clear, I'm not wondering why formatHeader takes a pointer to a 
> byte 
> > slice instead of a byte slice--indeed the itoa function above it does 
> this, 
> > and must do this--but rather why does it need the buffer parameter at 
> all, 
> > since it's a method and therefore already has access to its buffer. 
> > 
> > Is it simply to make the method more general, for possible future 
> changes, 
> > so that it can format headers using its flag and prefix settings but 
> write 
> > the header to some other buffer instead of its own?  Or is there a 
> deeper 
> > design reason that I'm missing? 
> > 
> > Please note that I'm not trying to criticize but rather trying to 
> > understand.  I often look at how code is written in the standard library 
> to 
> > improve my own code. 
>
> I think it's just a historical accident.  Originally Logger.Output 
> passed the address of a local variable to formatHeader.  Then in 
> https://golang.org/cl/4363045 Rob added a buf field to Logger, and 
> passed the address of that, with the CL description saying "rearrange 
> the code a little so we can avoid allocating a buffer on every call." 
> He could have changed formatHeader then to stop passing the buffer, 
> but he either didn't notice or decided not to mix that into the same 
> change. 
>
> Ian 
>

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [go-nuts] log package implementation question

[Ian Lance Taylor]

On Wed, Jun 7, 2017 at 11:12 AM,   wrote:
>
> I was perusing the code for the standard log package (src/log/log.go - The
> Go Programming Language) and was wondering why the formatHeader method (line
> 92) takes a pointer to byte slice since the only call to this method line
> (163) calls it with the address of its own byte slice member (and it only
> has one such member).
>
> To be clear, I'm not wondering why formatHeader takes a pointer to a byte
> slice instead of a byte slice--indeed the itoa function above it does this,
> and must do this--but rather why does it need the buffer parameter at all,
> since it's a method and therefore already has access to its buffer.
>
> Is it simply to make the method more general, for possible future changes,
> so that it can format headers using its flag and prefix settings but write
> the header to some other buffer instead of its own?  Or is there a deeper
> design reason that I'm missing?
>
> Please note that I'm not trying to criticize but rather trying to
> understand.  I often look at how code is written in the standard library to
> improve my own code.

I think it's just a historical accident.  Originally Logger.Output
passed the address of a local variable to formatHeader.  Then in
https://golang.org/cl/4363045 Rob added a buf field to Logger, and
passed the address of that, with the CL description saying "rearrange
the code a little so we can avoid allocating a buffer on every call."
He could have changed formatHeader then to stop passing the buffer,
but he either didn't notice or decided not to mix that into the same
change.

Ian

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[go-nuts] log package implementation question

[hanisch]

I was perusing the code for the standard log package (src/log/log.go - The 
Go Programming Language ) 
and was wondering why the formatHeader method (line 92) takes a pointer to 
byte slice since the only call to this method line (163) calls it with the 
address of its own byte slice member (and it only has one such member).

To be clear, I'm not wondering why formatHeader takes a pointer to a byte 
slice instead of a byte slice--indeed the itoa function above it does this, 
and must do this--but rather why does it need the buffer parameter at all, 
since it's a method and therefore already has access to its buffer.

Is it simply to make the method more general, for possible future changes, 
so that it can format headers using its flag and prefix settings but write 
the header to some other buffer instead of its own?  Or is there a deeper 
design reason that I'm missing?

Please note that I'm not trying to criticize but rather trying to 
understand.  I often look at how code is written in the standard library to 
improve my own code.

William

-- 
You received this message because you are subscribed to the Google Groups 
"golang-nuts" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to golang-nuts+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.