In CVS Emacs, calc-eval gives multiplication higher precidence than division

In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Francis Litterio]

In CVS Emacs, function calc-eval (in lisp/calc/calc.el) gives
multiplication higher precidence than division.  To reproduce, run
"emacs -q" and evaluate this form:

   (progn
     (require 'calc)
     (calc-eval "1 / 2 * 1000"))

The value of the above form is "5e-4" (as if operator '*' has higher
precidence than operator '/'), but it should be "500".
--
Fran

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Jay Belanger]

Francis Litterio <flitterio@gmail.com> writes:
...
> In CVS Emacs, function calc-eval (in lisp/calc/calc.el) gives
> multiplication higher precidence than division.

That's not a bug.  While many systems give division a higher precendence
than multiplication, that isn't a universal rule.  Calc has always given
multiplication a higher precendence, and something like
1 / 2 * 1000
can be viewed as shorthand for
   1
--------
2 * 1000
Some people don't like this, some people do, but this choice is
established Calc behavior and is highlighted in the manual.  To divide
first, it's probably better style regardless to put in the parentheses:
(1 / 2) * 1000

Jay

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Francis Litterio]

Jay Belanger wrote:

> Francis Litterio writes:
> ...
>> In CVS Emacs, function calc-eval (in lisp/calc/calc.el) gives
>> multiplication higher precidence than division.
>
> That's not a bug.  While many systems give division a higher precendence
> than multiplication, that isn't a universal rule.  Calc has always given
> multiplication a higher precendence, and something like
> 1 / 2 * 1000
> can be viewed as shorthand for
>    1
> --------
> 2 * 1000
> Some people don't like this, some people do, but this choice is
> established Calc behavior and is highlighted in the manual.  To divide
> first, it's probably better style regardless to put in the parentheses:
> (1 / 2) * 1000

Thanks, Jay.

My apologies for rehashing old news.  I just discovered the discussion
thread at:

	http://thread.gmane.org/gmane.emacs.devel/69436
--
Fran

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Stefan Monnier]

> That's not a bug.  While many systems give division a higher precendence
> than multiplication, that isn't a universal rule.

Actually, I don't know of any system that gives higher precedence to
division.  The general rule that (AFAIK) all programming languages
follow (other than those that don't use infix-with-precedence, such as
Lisp and Smalltalk) is that * and / have the same precedence (and are
left associative).

I think Calc is plainly wrong is giving higher precedence to the
multiplication because it's completely non standard.  But since I don't
use Calc much and since it's been that way for a while, I won't fight
for it.


        Stefan

Re: bug#920: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
> I think Calc is plainly wrong is giving higher precedence to the
> multiplication because it's completely non standard.  But since I don't
> use Calc much and since it's been that way for a while, I won't fight
> for it.


I think there are good reasons fixing it:

- There are enough people out there that are not aware of the precedence
rules. Making calc follow them would help remove that confusion. (And I
would feel safer when crossing bridges.)
- Not following standards makes Emacs harder to use for new users.

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Randal L. Schwartz]

>>>>> "Stefan" == Stefan Monnier <monnier@iro.umontreal.ca> writes:

Stefan> I think Calc is plainly wrong is giving higher precedence to the
Stefan> multiplication because it's completely non standard.  But since I don't
Stefan> use Calc much and since it's been that way for a while, I won't fight
Stefan> for it.

Given the original author's mathematics bias, it made sense (to me, as
an early tester and contributor) that:

   2 * 3 / 4 * 5

would be parsed as

   (2 * 3) / (4 * 5)

in a mathematical sense.

Admittedly, if you're a computer-science guy and not a math guy, this might
seem initially odd.

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.vox.com/ for Smalltalk and Seaside discussion

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Carsten Dominik]

On Sep 7, 2008, at 10:32 PM, Randal L. Schwartz wrote:

>>>>>> "Stefan" == Stefan Monnier <monnier@iro.umontreal.ca> writes:
>
> Stefan> I think Calc is plainly wrong is giving higher precedence to  
> the
> Stefan> multiplication because it's completely non standard.  But  
> since I don't
> Stefan> use Calc much and since it's been that way for a while, I  
> won't fight
> Stefan> for it.


I would fight for the current calc implementation, it is an important  
convenience when entering equations.  I am relying on it every day.

- Carsten



>
>
> Given the original author's mathematics bias, it made sense (to me, as
> an early tester and contributor) that:
>
>   2 * 3 / 4 * 5
>
> would be parsed as
>
>   (2 * 3) / (4 * 5)
>
> in a mathematical sense.
>
> Admittedly, if you're a computer-science guy and not a math guy,  
> this might
> seem initially odd.
>
> -- 
> Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503  
> 777 0095
> <merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
> Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
> See http://methodsandmessages.vox.com/ for Smalltalk and Seaside  
> discussion
>
>
>

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Lennart Borgman]

On 9/8/08, Carsten Dominik <carsten.dominik@gmail.com> > I would fight
for the current calc implementation, it is an important
> convenience when entering equations.  I am relying on it every day.

Exactly how are you relying on that the current behaviour is the default?

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Carsten Dominik]

On Sep 8, 2008, at 1:01 PM, Lennart Borgman wrote:

> On 9/8/08, Carsten Dominik <carsten.dominik@gmail.com> > I would fight
> for the current calc implementation, it is an important
>> convenience when entering equations.  I am relying on it every day.
>
> Exactly how are you relying on that the current behaviour is the  
> default?


By typing algebraic equations into the minibuffer and saving tons of  
key strokes by leaving off parenthesis.  I often type equations like

  p = n k T / mu mp

and it saves me 4 key strokes to not having to type the parenthesis

  p = (n k T) / (mu mp)

- Carsten

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Lennart Borgman (gmail)]

Carsten Dominik wrote:
> 
> On Sep 8, 2008, at 1:01 PM, Lennart Borgman wrote:
> 
>> On 9/8/08, Carsten Dominik <carsten.dominik@gmail.com> > I would fight
>> for the current calc implementation, it is an important
>>> convenience when entering equations.  I am relying on it every day.
>>
>> Exactly how are you relying on that the current behaviour is the default?
> 
> 
> By typing algebraic equations into the minibuffer and saving tons of key
> strokes by leaving off parenthesis.  I often type equations like
> 
>  p = n k T / mu mp
> 
> and it saves me 4 key strokes to not having to type the parenthesis
> 
>  p = (n k T) / (mu mp)


Yes, sorry, I should not have asked like that. I get a little bit upset
that many people here accept defaults that can make it harder for
newbees just for a little bit convenience. (But I guess I am a bit
unfair. Someone more in favour of the view opposite to mine in this
question can surely put it more nicely.)

Of course I can see that the behaviour you want saves you a lot of key
strokes. My question is really about the default value. You can easily
change that, of course. The question is really if you prefer the old
default and in that case what arguments you want to present for that.

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Randal L. Schwartz]

>>>>> "Lennart" == Lennart Borgman (gmail) <lennart.borgman@gmail.com> writes:

Lennart> Of course I can see that the behaviour you want saves you a lot of key
Lennart> strokes. My question is really about the default value. You can easily
Lennart> change that, of course. The question is really if you prefer the old
Lennart> default and in that case what arguments you want to present for that.

There's no need to change the longstanding defaults, because it would break
things.  You already have a knob you can tweak if you think that "2 a / 3 b"
should multiply by b, rather than divide by it.  That's a personal choice.

Frankly, I'd never touch that knob, because it looks to me like it should
divide by b.  And I think most mathemeticians would agree, and that's the
target for that part of Calc.  Stop thinking computer science.  Think
*math* notation.

-- 
Randal L. Schwartz - Stonehenge Consulting Services, Inc. - +1 503 777 0095
<merlyn@stonehenge.com> <URL:http://www.stonehenge.com/merlyn/>
Smalltalk/Perl/Unix consulting, Technical writing, Comedy, etc. etc.
See http://methodsandmessages.vox.com/ for Smalltalk and Seaside discussion

bug#914: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Lennart Borgman (gmail)]

Randal L. Schwartz wrote:
>>>>>> "Lennart" == Lennart Borgman (gmail) <lennart.borgman@gmail.com> writes:
> 
> Lennart> Of course I can see that the behaviour you want saves you a lot of key
> Lennart> strokes. My question is really about the default value. You can easily
> Lennart> change that, of course. The question is really if you prefer the old
> Lennart> default and in that case what arguments you want to present for that.
> 
> There's no need to change the longstanding defaults, because it would break
> things.  You already have a knob you can tweak if you think that "2 a / 3 b"
> should multiply by b, rather than divide by it.  That's a personal choice.

I think communication can be very difficult if you have too much
personal choice in it. Math is about communication (for a special class
of logical problems).

The original bug report was about this

  (calc-eval "1 / 2 * 1000")

Calc treats this as 1 / (2 * 1000) which is just plain wrong. I have
never seen a mathematician treat it that way - though I have seen many
pupils doing that.


> Frankly, I'd never touch that knob, because it looks to me like it should
> divide by b.  And I think most mathemeticians would agree, and that's the
> target for that part of Calc.  Stop thinking computer science.  Think
> *math* notation.

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Stefan Monnier]

Lennart> Of course I can see that the behaviour you want saves you a lot of key
Lennart> strokes. My question is really about the default value. You can easily
Lennart> change that, of course. The question is really if you prefer the old
Lennart> default and in that case what arguments you want to present for that.

> There's no need to change the longstanding defaults, because it would break
> things.  You already have a knob you can tweak if you think that "2 a / 3 b"
> should multiply by b, rather than divide by it.  That's a personal choice.

Nobody in this thread has said that "2 a / 3 b" should multiply by b.
What I (and others) have said is that "2 * a / 3 * b" should multiply by b.


        Stefan

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Richard M. Stallman]

    There's no need to change the longstanding defaults, because it would break
    things.  You already have a knob you can tweak if you think that "2 a / 3 b"
    should multiply by b, rather than divide by it.  That's a personal choice.

    Frankly, I'd never touch that knob, because it looks to me like it should
    divide by b.

I agree completely -- in that case, it is natural for b to go in the
denominator.

The same is true for 1 / 2 * 1000: the meaning of 1/2000 is more
natural, mathematically.

What Calc does now is clearly correct.

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Stefan Monnier]

> What Calc does now is clearly correct.

And we clearly disagree.


        Stefan

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Stefan Monnier]

> I would fight for the current calc implementation, it is an important
> convenience when entering equations.  I am relying on it every day.

It's really a very minor issue that only determines when and where you
need to place parentheses.  And you can control it with a config var in
case you don't like the default.


        Stefan

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Jay Belanger]

>> That's not a bug.  While many systems give division a higher precendence
>> than multiplication, that isn't a universal rule.
>
> Actually, I don't know of any system that gives higher precedence to
> division.

According to Wikipedia:
  An expression like 1/2x is interpreted as 1/(2x) by TI-82, but as
  (1/2)x by TI-83. While the first interpretation may be expected by
  some users, only the latter is in agreement with the standard rules
  stated above. 

> The general rule that (AFAIK) all programming languages
> follow (other than those that don't use infix-with-precedence, such as
> Lisp and Smalltalk) is that * and / have the same precedence (and are
> left associative).

That behavior is available, but you have to ask for it by setting
`calc-multiplication-has-precedence' to nil.

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Stefan Monnier]

>>> That's not a bug.  While many systems give division a higher precendence
>>> than multiplication, that isn't a universal rule.
>> Actually, I don't know of any system that gives higher precedence to
>> division.

> According to Wikipedia:
>   An expression like 1/2x is interpreted as 1/(2x) by TI-82, but as
>   (1/2)x by TI-83. While the first interpretation may be expected by
>   some users, only the latter is in agreement with the standard rules
>   stated above. 

That could be due to a higher precedence, but could also be due to an
equal precedence together with left-associativity.  Given that (a*b)/c
is often mathematically equal to a*(b/c), the difference is generally
hard to see, unless rounding comes into play.

All textbooks and language references I know use the "equal precedence
and left associativity" rule rather than giving higher precedence to
the division.

>> The general rule that (AFAIK) all programming languages
>> follow (other than those that don't use infix-with-precedence, such as
>> Lisp and Smalltalk) is that * and / have the same precedence (and are
>> left associative).

> That behavior is available, but you have to ask for it by setting
> `calc-multiplication-has-precedence' to nil.

I know.  But we're talking about the default behavior.


        Stefan

Re: In CVS Emacs, calc-eval gives multiplication higher precidence than division

[Lennart Borgman (gmail)]

Stefan Monnier wrote:
>> According to Wikipedia:
>>   An expression like 1/2x is interpreted as 1/(2x) by TI-82, but as
>>   (1/2)x by TI-83. While the first interpretation may be expected by
>>   some users, only the latter is in agreement with the standard rules
>>   stated above. 
> 
> That could be due to a higher precedence, but could also be due to an
> equal precedence together with left-associativity.  Given that (a*b)/c
> is often mathematically equal to a*(b/c), the difference is generally
> hard to see, unless rounding comes into play.

But that is a bit different from (a/b)*c, a/(b*c) ... - otherwise I agree.