[9fans] A new language for Plan 9

[9fans] A new language for Plan 9

[Pietro Gagliardi]

Hello. I started working on Bentley, a new programming language. This
was inspired by and is based on the pseudocode in Jon Bentley's
"Programming Pearls" - a column for the CACM that became a book. The
compiler generates Assembly in a temporary file, then calls up the
assembler to make the program. Currently, the assembly that can be
generated is a subset of the language that can be parsed, but I'm
working on it.

	% rc /n/sources/contrib/pietro/bentley.bundle
	...
	% mk
	...
	% 8b asm.b
	% 8l -o asm asm.8
	% asm
	5
	1
	% cat test.b # for what the parser is capable of

The idea is simple: indentation as block style. Instead of

	if (key == 'c') {
		scanline();
		runcommand();
	} else {
		generate(key);
		assemble(key);
	}

one does

	if key = 'c' then
			scanline
			runcommand
		else
			generate(key)
			assemble(key)

This is similar to Python, and prevents the nesting ambiguity of C,
Pascal, and some other languages that use block delimiters.

Another feature I hope to supply is bit arrays.

	x: 32 bits
	x := 406 { a bit array is stored atomically }
	x[2] ;= 0 { and yet can be used structurally }

On the x86 platform, the BTS, BTC, and BT instructions facilitate
this. I don't know their equivalents on other processors. This feature
is completely independent of endianness if I use the instructions.

Bentley also will have nested functions, a Pascal-like for statement
(with variable steps instead of 1/-1), and a loop statement for
infinite loops.

To ensure programmers will use good style, Bentley will lack goto. To
break out of nested loops, you can use the breakout statement.

Finally, there will be two modes: hosted and standalone. The
standalone keyword changes this. Hosted mode can access print to
stdout and stderr, read from stdin, new, renew (like realloc), delete,
and a string type.

Pietro

Re: [9fans] A new language for Plan 9

[erik quanstrom]

> one does
>
> 	if key = 'c' then
> 			scanline
> 			runcommand
> 		else
> 			generate(key)
> 			assemble(key)
>
> This is similar to Python, and prevents the nesting ambiguity of C,
> Pascal, and some other languages that use block delimiters.

don't forget fortran.

- erik

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

On May 1, 2008, at 9:26 PM, erik quanstrom wrote:

>> one does
>>
>> 	if key = 'c' then
>> 			scanline
>> 			runcommand
>> 		else
>> 			generate(key)
>> 			assemble(key)
>>
>> This is similar to Python, and prevents the nesting ambiguity of C,
>> Pascal, and some other languages that use block delimiters.
>
> don't forget fortran.
>
[off topic] I wonder why it took 20 or so years for Fortran to
introduce IF..ELSE..END IF. Probably after Kernighan and Plauger,
1974, it finally came to Backus' senses. :-P

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

On May 1, 2008, at 10:21 PM, Pietro Gagliardi wrote:

>
> On May 1, 2008, at 9:26 PM, erik quanstrom wrote:
>
>>> one does
>>>
>>> 	if key = 'c' then
>>> 			scanline
>>> 			runcommand
>>> 		else
>>> 			generate(key)
>>> 			assemble(key)
>>>
>>> This is similar to Python, and prevents the nesting ambiguity of C,
>>> Pascal, and some other languages that use block delimiters.
>>
>> don't forget fortran.
>>
> [off topic] I wonder why it took 20 or so years for Fortran to
> introduce IF..ELSE..END IF. Probably after Kernighan and Plauger,
> 1974, it finally came to Backus' senses. :-P
>
Also, the decision not to include goto satisfies some of the elements
they describe, as well as being deliberate. There are good languages
that lack gotos completely and yet you can write programs in it. Take
Java as a perfect example - it reserves goto but doesn't use it. Java
has been used for production-class and enterprise-class applications
(J2EE stands for Java 2 Enterprise Edition, after all). In Bentley you
can say

	goto := 4

Re: [9fans] A new language for Plan 9

[Matt Erickson]

On 2008-05-02, erik quanstrom <quanstro@quanstro.net> pondered onto the tubes:
>> one does
>>
>> 	if key = 'c' then
>> 			scanline
>> 			runcommand
>> 		else
>> 			generate(key)
>> 			assemble(key)
>>
>> This is similar to Python, and prevents the nesting ambiguity of C,
>> Pascal, and some other languages that use block delimiters.
>
> don't forget fortran.

I try to, oh God how I try to.  Then I come back to work in the
morning, and have to remember it all over again.

--
Matt Erickson           <peawee@peawee.net>
Political correctness is not insane. You want to call it "alternatively
reality-based". - Richard Bos, Usenet

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

On May 7, 2008, at 5:24 AM, Matt Erickson wrote:

> On 2008-05-02, erik quanstrom <quanstro@quanstro.net> pondered onto
> the tubes:
>>> one does
>>>
>>> 	if key = 'c' then
>>> 			scanline
>>> 			runcommand
>>> 		else
>>> 			generate(key)
>>> 			assemble(key)
>>>
>>> This is similar to Python, and prevents the nesting ambiguity of C,
>>> Pascal, and some other languages that use block delimiters.
>>
>> don't forget fortran.
>
> I try to, oh God how I try to.  Then I come back to work in the
> morning, and have to remember it all over again.
>
> --
> Matt Erickson           <peawee@peawee.net>
> Political correctness is not insane. You want to call it
> "alternatively
> reality-based". - Richard Bos, Usenet
>

Good luck. My next compiler will do Fortran 77, and hopefully 90. It
will be #f, where # is the prefix (so 8f for 386).

A new Bentley is up, in the usual place. This adds full procedures and
functions with arguments, expression statements (like in C), and
demotes assignment from a statement to an operator. The next version
will have it right-associative (a := b := c := 0). I also plan to have
scope.

I'm surprised at how quickly I'm getting through this. I expected some
2 year project getting as much as the Sieve of Erastosthenes to work.

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

A new Bentley is up. This defines hopefully all of the bitwise
operators I will support:

	& | ^ << >> ~ (as in C)
	<@ >@ (roll left and roll right, also called circular shift)

I'm still working on the base language, getting type checking working
and allowing logical AND, OR, XOR, NOT. But perhaps the hardest part
is writing the manual for this language.

Re: [9fans] A new language for Plan 9

[Rob Pike]

Indentation by white space is a very bad idea in my experience.
Superficially attractive but ultimately very dangerous. I once spent a
couple of days tracking down a bug caused by a source-to-source code
tool that broke a major program because the code it was injecting into
had indented one more space, causing the injecting code to break the
control flow. It was nearly impossible to track down.

I have lots of other examples of lesser disasters. As code grows,
white space indentation becomes ever more problematic. It's a
maintenance disaster.

Put it this way: It's unwise to make program structure depend on
invisible characters.

-rob

Re: [9fans] A new language for Plan 9

[Federico G. Benavento]

> Put it this way: It's unwise to make program structure depend on
> invisible characters.

a white space is something hard to find, some time ago I helped a friend
who couldn't get a mkfile working, he got something like:
"mk: mkfile:6: syntax error; expected one of :<="
all due to a ' ' in what was supposed to be an empty line.

Federico G. Benavento

---
/bin/fortune:
You are in a maze of twisty little passages, all alike.

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

On May 1, 2008, at 9:12 PM, Federico G. Benavento wrote:

>> Put it this way: It's unwise to make program structure depend on
>> invisible characters.
>
> a white space is something hard to find, some time ago I helped a
> friend
> who couldn't get a mkfile working, he got something like:
> "mk: mkfile:6: syntax error; expected one of :<="
> all due to a ' ' in what was supposed to be an empty line.
>
> Federico G. Benavento
>
Another thing: lines with only whitespace are naturally ignored (as
far as I know - I am human) and don't break the block structure.

The space in an empty line should be a bug, right?

Re: [9fans] A new language for Plan 9

[Rob Pike]

(By 'indentation' of course I mean 'indentation to define structure')

-rob

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

On May 1, 2008, at 9:52 PM, Rob Pike wrote:

> Indentation by white space is a very bad idea in my experience.
> Superficially attractive but ultimately very dangerous. I once spent a
> couple of days tracking down a bug caused by a source-to-source code
> tool that broke a major program because the code it was injecting into
> had indented one more space, causing the injecting code to break the
> control flow. It was nearly impossible to track down.
>
This scenario is why I decided to require hard tabs as indents. You
can't use a space. But a space halts the indentation, so the problem
is still there. The next version will just ignore spaces, and treat
the first non-whitespace character as the beginning of a line.

> I have lots of other examples of lesser disasters. As code grows,
> white space indentation becomes ever more problematic. It's a
> maintenance disaster.
>
> Put it this way: It's unwise to make program structure depend on
> invisible characters.
There's a language made entirely of said invisible characters, called
Whitespace. It's esoteric, but it works. And Python, which has the
same style, is a phenomenal success. Whether or not indentation works
relies on the programmer.

Re: [9fans] A new language for Plan 9

[Robert William Fuller]

Pietro Gagliardi wrote:

<snip>

>> Put it this way: It's unwise to make program structure depend on
>> invisible characters.
> There's a language made entirely of said invisible characters, called
> Whitespace. It's esoteric, but it works. And Python, which has the same
> style, is a phenomenal success. Whether or not indentation works relies
> on the programmer.

I don't use Python for this very reason.  This is probably why Ruby
exists.  I will not use your language for the same reason.  By adopting
such draconian white space rules you automatically alienate a large
number of programmers.

I consider this one of the larger mistakes in programming language
design, akin to making the period the most important token in the COBOL
language, or using whitespace as a separator in FORTRAN (which
incidentally lost NASA a space probe.)

Re: [9fans] A new language for Plan 9

[Chad Dougherty]

Robert William Fuller wrote:
> I don't use Python for this very reason.  This is probably why Ruby
> exists.  I will not use your language for the same reason.  By adopting
> such draconian white space rules you automatically alienate a large
> number of programmers.
>

A blind programmer once told me that Python's whitespace block structure
was simply too high of a barrier for him to use it.

	-Chad

Re: [9fans] A new language for Plan 9

[Robert William Fuller]

Chad Dougherty wrote:
> Robert William Fuller wrote:
>> I don't use Python for this very reason.  This is probably why Ruby
>> exists.  I will not use your language for the same reason.  By
>> adopting such draconian white space rules you automatically alienate a
>> large number of programmers.
>>
>
> A blind programmer once told me that Python's whitespace block structure
> was simply too high of a barrier for him to use it.

Gosh and I thought being blind was a pre-requisite for programming in
Python.  Thanks for clearing that up.

Re: [9fans] A new language for Plan 9

[David Arnold]

On 02/05/2008, at 8:34 AM, Chad Dougherty wrote:

> Robert William Fuller wrote:
>> I don't use Python for this very reason.  This is probably why
>> Ruby exists.  I will not use your language for the same reason.
>> By adopting such draconian white space rules you automatically
>> alienate a large number of programmers.
>
> A blind programmer once told me that Python's whitespace block
> structure was simply too high of a barrier for him to use it.

straying off-topic, but ...

the Python distribution includes a tool called 'pindent'.  it happily
annotates Python source with block-closing comments and converts
haphazardly indented source with block-closing comments into
correctly indented Python source.

    http://svn.python.org/projects/python/trunk/Tools/scripts/pindent.py




d

Re: [9fans] A new language for Plan 9

[lucio]

> the Python distribution includes a tool called 'pindent'.  it happily
> annotates Python source with block-closing comments and converts
> haphazardly indented source with block-closing comments into
> correctly indented Python source.

This is irony in a league all of its own!

++L

Re: [9fans] A new language for Plan 9

[Skip Tavakkolian]

my original suggestion for ratpy wasn't taken seriously, so i'll propose it again:

http://groups.google.com/group/comp.os.plan9/msg/29eb245edcb78e91

>>> I don't use Python for this very reason.  This is probably why
>>> Ruby exists.  I will not use your language for the same reason.
>>> By adopting such draconian white space rules you automatically
>>> alienate a large number of programmers.
>>
>> A blind programmer once told me that Python's whitespace block
>> structure was simply too high of a barrier for him to use it.
>
> straying off-topic, but ...
>
> the Python distribution includes a tool called 'pindent'.  it happily
> annotates Python source with block-closing comments and converts
> haphazardly indented source with block-closing comments into
> correctly indented Python source.
>
>     http://svn.python.org/projects/python/trunk/Tools/scripts/pindent.py

Re: [9fans] A new language for Plan 9

[Bruce Ellis]

ratpie! tasty. i thought a pindent was what a pinhead gets when you
scone him with a frypan.

brucee

On Sun, May 4, 2008 at 9:28 AM, Skip Tavakkolian <9nut@9netics.com> wrote:
> my original suggestion for ratpy wasn't taken seriously, so i'll propose it again:
>
> http://groups.google.com/group/comp.os.plan9/msg/29eb245edcb78e91
>
> >>> I don't use Python for this very reason.  This is probably why
> >>> Ruby exists.  I will not use your language for the same reason.
> >>> By adopting such draconian white space rules you automatically
> >>> alienate a large number of programmers.
> >>
> >> A blind programmer once told me that Python's whitespace block
> >> structure was simply too high of a barrier for him to use it.
> >
> > straying off-topic, but ...
> >
> > the Python distribution includes a tool called 'pindent'.  it happily
> > annotates Python source with block-closing comments and converts
> > haphazardly indented source with block-closing comments into
> > correctly indented Python source.
> >
> >     http://svn.python.org/projects/python/trunk/Tools/scripts/pindent.py
>
>
>

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

We have two ratpies. They're ruby and perl. Pick your poison.

On May 3, 2008, at 7:32 PM, Bruce Ellis wrote:

> ratpie! tasty. i thought a pindent was what a pinhead gets when you
> scone him with a frypan.
>
> brucee
>
> On Sun, May 4, 2008 at 9:28 AM, Skip Tavakkolian <9nut@9netics.com>
> wrote:
>> my original suggestion for ratpy wasn't taken seriously, so i'll
>> propose it again:
>>
>> http://groups.google.com/group/comp.os.plan9/msg/29eb245edcb78e91
>>
>>>>> I don't use Python for this very reason.  This is probably why
>>>>> Ruby exists.  I will not use your language for the same reason.
>>>>> By adopting such draconian white space rules you automatically
>>>>> alienate a large number of programmers.
>>>>
>>>> A blind programmer once told me that Python's whitespace block
>>>> structure was simply too high of a barrier for him to use it.
>>>
>>> straying off-topic, but ...
>>>
>>> the Python distribution includes a tool called 'pindent'.  it
>>> happily
>>> annotates Python source with block-closing comments and converts
>>> haphazardly indented source with block-closing comments into
>>> correctly indented Python source.
>>>
>>>    http://svn.python.org/projects/python/trunk/Tools/scripts/pindent.py
>>
>>
>>
>

Re: [9fans] A new language for Plan 9

[Eric Van Hensbergen]

On Sat, May 3, 2008 at 6:32 PM, Bruce Ellis <bruce.ellis@gmail.com> wrote:
> ratpie! tasty. i thought a pindent was what a pinhead gets when you
>  scone him with a frypan.
>

Can I have a piece without so much rat in it?

   -eric

"Its the Bishop of Liechester"

Re: [9fans] A new language for Plan 9

[Steve Simon]

> Can I have a piece without so much rat in it?

I had stir fried rat in vietnam once - well you gotta try things,
tasted a bit like wild (strong chewy) chicken.

-Steve

Re: [9fans] A new language for Plan 9

[Bruce Ellis]

Well KenC doesn't have any rat with it.

So you want spam spam spam bacon and KenC.

Or python and a bucket.

brucee

On Sun, May 4, 2008 at 12:43 PM, Eric Van Hensbergen <ericvh@gmail.com> wrote:
> On Sat, May 3, 2008 at 6:32 PM, Bruce Ellis <bruce.ellis@gmail.com> wrote:
> > ratpie! tasty. i thought a pindent was what a pinhead gets when you
> >  scone him with a frypan.
> >
>
> Can I have a piece without so much rat in it?
>
>   -eric
>
> "Its the Bishop of Liechester"

Re: [9fans] A new language for Plan 9

[John Stalker]

Just out of curiosity, why did alef die, or are some of you
still using it?

--
John Stalker
School of Mathematics
Trinity College Dublin
tel +353 1 896 1983
fax +353 1 896 2282

Re: [9fans] A new language for Plan 9

[Steve Simon]

As I understand it the load of supporting two libraries was too much work, also
the alef (and perhaps limbo) experience lead to libthread which provides much of
the same functionality - abet not quite as neatly.

the sources of the 2nd edition alef have been released and there was a
one line change to make the code work on the 3rd (and I assume 4th) editions -
the memory layout of processes has changed a little and the alef library needs
to know (the stack address I think), for the fix see 9fans passim.

-Steve

Re: [9fans] A new language for Plan 9

[lucio]

> Just out of curiosity, why did alef die, or are some of you
> still using it?

A victim of rationalisation and reality: maintaining separate
libraries for two languages is a luxury Plan 9 colud no longer afford,
if I understood the rationale at the time.

++L

Re: [9fans] A new language for Plan 9

[Bruce Ellis]

I think the quote of the day was "We already support one C-like language".

brucee

On Sun, May 4, 2008 at 5:57 PM,  <lucio@proxima.alt.za> wrote:
> > Just out of curiosity, why did alef die, or are some of you
> > still using it?
>
> A victim of rationalisation and reality: maintaining separate
> libraries for two languages is a luxury Plan 9 colud no longer afford,
> if I understood the rationale at the time.
>
> ++L

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

s/one/two/ # don't forget c++

On May 4, 2008, at 8:18 AM, Bruce Ellis wrote:

> I think the quote of the day was "We already support one C-like
> language".
>
> brucee
>
> On Sun, May 4, 2008 at 5:57 PM,  <lucio@proxima.alt.za> wrote:
>>> Just out of curiosity, why did alef die, or are some of you
>>> still using it?
>>
>> A victim of rationalisation and reality: maintaining separate
>> libraries for two languages is a luxury Plan 9 colud no longer
>> afford,
>> if I understood the rationale at the time.
>>
>> ++L
>

Re: [9fans] A new language for Plan 9

[Eric Van Hensbergen]

On Sun, May 4, 2008 at 8:48 AM, Pietro Gagliardi <pietro10@mac.com> wrote:
> s/one/two/ # don't forget c++
>

If only we could....

          -eric

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

I put up a new Bentley. This has support for, hopefully, all control
structures:

	if e then					if..then
		s
	if e then					if..then..else
			s
		else
			s
	while e do				while..do
		s
	loop					loop (infinite loop)
		s
	repeat					repeat..until (like do..while(!))
			s
		until e
	for i := e to e [by e] do			like in pascal, but with steps
		s
	for i := e downto e [by e] do	like in pascal, but with steps
		s

The file asm.b is a simple demonstration of the last two.

The Bentley compiler, when built, is named #b, so the 386 compiler
(and the only one so far) is 8b. The name is changed on build. This is
used exactly like the C compilers:

	8b asm.b
	8l -o asm asm.8
	asm

The -S option produces an Assembly listing, but does not build (the C
compilers do).

My only regret so far is that for loops generate a medium sized piece
of spaghetti (yum). See the comment in 8.c for details.

Set for the next release: arguments in procedures, functions, a more
sophisticated stack size algorithm, bit arrays.

Re: [9fans] A new language for Plan 9

[andrey mirtchovski]

>  Set for the next release: bit arrays.

"I'd use plan 9 before i'd use bitfields" -- pjw

Re: [9fans] A new language for Plan 9

[erik quanstrom]

> >  Set for the next release: bit arrays.
>
> "I'd use plan 9 before i'd use bitfields" -- pjw
>

funny, i've only know him as a bit array.

- erik

Re: [9fans] A new language for Plan 9

[Pietro Gagliardi]

On May 4, 2008, at 10:24 PM, andrey mirtchovski wrote:

>> Set for the next release: bit arrays.
>
> "I'd use plan 9 before i'd use bitfields" -- pjw
>
I'm aware of that quote. The design of bit arrays will be totally
different from the design of bitfields. They are true arrays that can
be used in an implementation-independent manner. Given

	b: 32 bits
	b[0] := 4

the compiler will determine whether b[0] is at the far left or far
right of the binary version of that value. This means you can have
code, theoretically, like

	function getlong: 32 bits
		b: 32 bits
		b[31..24] := getchar - '0'
		b[23..16] := getchar - '0'
		b[15..8] := getchar - '0'
		b[7..0] := getchar - '0'
		return b

and then use getlong as follows:

	x: 32 bits = getlong
	if x = 400 then
		{ ... }

This is a severe blow to bit fields.

No, I'm not doing a school assignment. This is more of a hobby project.

Re: [9fans] A new language for Plan 9

[Bruce Ellis]

wow, sounds like a freshman assignment. don't steal ".b" from limbo, use limbo.

brucee

On Mon, May 5, 2008 at 11:45 AM, Pietro Gagliardi <pietro10@mac.com> wrote:
> I put up a new Bentley. This has support for, hopefully, all control
> structures:
>
>        if e then                                       if..then
>                s
>        if e then                                       if..then..else
>                        s
>                else
>                        s
>        while e do                              while..do
>                s
>        loop                                    loop (infinite loop)
>                s
>        repeat                                  repeat..until (like
> do..while(!))
>                        s
>                until e
>        for i := e to e [by e] do                       like in pascal, but
> with steps
>                s
>        for i := e downto e [by e] do   like in pascal, but with steps
>                s
>
> The file asm.b is a simple demonstration of the last two.
>
> The Bentley compiler, when built, is named #b, so the 386 compiler (and the
> only one so far) is 8b. The name is changed on build. This is used exactly
> like the C compilers:
>
>        8b asm.b
>        8l -o asm asm.8
>        asm
>
> The -S option produces an Assembly listing, but does not build (the C
> compilers do).
>
> My only regret so far is that for loops generate a medium sized piece of
> spaghetti (yum). See the comment in 8.c for details.
>
> Set for the next release: arguments in procedures, functions, a more
> sophisticated stack size algorithm, bit arrays.
>
>
>

Re: [9fans] A new language for Plan 9

[ron minnich]

you can do what you will, with your indentation-based language, but
that won't change the fact that indentation for lexical scope is a
horrible idea.

I first saw it in a language in 1978 called Offal, by Aron Insinga.
Aron was smart: after 6 weeks, he said, "this sucks", and put it away.
When I saw the beginnings of Python 10 years or so later I thought: "
I wonder when they'll figure it out". 20 years later, they haven't.
That says something.

The other ideas are interesting. Don't wreck yourself on the bad idea
of indenting.

Popularity is no measure of goodness. Or have you used Windows lately :-)

ron

Re: [9fans] A new language for Plan 9

[John Barham]

>  I first saw it in a language in 1978 called Offal, by Aron Insinga.

Well with a name like Offal at least he wasn't setting expectations too high...

Re: [9fans] A new language for Plan 9

[ron minnich]

On Thu, May 1, 2008 at 9:41 PM, John Barham <jbarham@gmail.com> wrote:
> >  I first saw it in a language in 1978 called Offal, by Aron Insinga.
>
>  Well with a name like Offal at least he wasn't setting expectations too high...
>
>

Just about as high as Python went, it turns out :-)

ron

Re: [9fans] A new language for Plan 9

[John Barham]

On Thu, May 1, 2008 at 9:54 PM, ron minnich <rminnich@gmail.com> wrote:
> On Thu, May 1, 2008 at 9:41 PM, John Barham <jbarham@gmail.com> wrote:
>  > >  I first saw it in a language in 1978 called Offal, by Aron Insinga.
>  >
>  >  Well with a name like Offal at least he wasn't setting expectations too high...
>  >
>
>  Just about as high as Python went, it turns out :-)

Touché.  At least it wasn't called Parrot or it might have been dead
from the start.

  John

Re: [9fans] A new language for Plan 9

[John Barham]

Rob Pike wrote:

>  I have lots of other examples of lesser disasters. As code grows,
>  white space indentation becomes ever more problematic. It's a
>  maintenance disaster.

I beg to differ, at least when it comes to my experience working w/
Python.  I work day in and day out on a 50,000+ line Python
application and can't recall a single bug that was caused by
whitespace indentation.  But granted Python is better hand-written
than generated...

Working on a constantly moving code-base I've experienced more
problems due to Python's dynamic typing, but that's a double-edged
sword and on balance it's still a big plus.

  John

Re: [9fans] A new language for Plan 9

[andrey mirtchovski]

> Indentation by white space is a very bad idea in my experience.

it could just be possible that you're using an editor that is not
aware of the particular indentation requirements of said language, no?

does it, at least, implement color coding?

:D

Re: [9fans] A new language for Plan 9

[erik quanstrom]

> it could just be possible that you're using an editor that is not
> aware of the particular indentation requirements of said language, no?
>
> does it, at least, implement color coding?

you wascale wabble wowser!  you're won of those pwython wuffians, awren't
you?

- erik

Re: [9fans] A new language for Plan 9

[John Barham]

Pietro Gagliardi wrote:

> The compiler generates Assembly in a temporary file,
> then calls up the assembler to make the program.

That sounds somewhat similar to Dan Bernstein's qhasm
(http://cr.yp.to/qhasm.html) which is a semi-portable assembly
language combining C-like syntax w/ direct access to registers.
Anathema to the Plan 9 philosophy I suppose but given that clock
speeds seem to have hit a wall it's one way to wring out more speed.
To be fair he seems to intend it to be used only for number crunching
inner loops rather than as a general purpose language.

  John

Re: [9fans] A new language for Plan 9

[erik quanstrom]

> That sounds somewhat similar to Dan Bernstein's qhasm
> (http://cr.yp.to/qhasm.html) which is a semi-portable assembly
> language combining C-like syntax w/ direct access to registers.
> Anathema to the Plan 9 philosophy I suppose but given that clock
> speeds seem to have hit a wall it's one way to wring out more speed.
> To be fair he seems to intend it to be used only for number crunching
> inner loops rather than as a general purpose language.

issues per clock have been rising on intel chips.  ibm power chips are
shipping at 5Ghz.  for typical parallel number-crunching app, i would
think that performance is increasing by leaps and bounds with the
addition of more cores.

- erik

Re: [9fans] A new language for Plan 9

[John Stalker]

> The idea is simple: indentation as block style.

Religious matter.  Do as you wish, but expect flames.

> Another feature I hope to supply is bit arrays.

ok

> Bentley also will have nested functions, a Pascal-like for statement
> (with variable steps instead of 1/-1), and a loop statement for
> infinite loops.

All languages have infinite loops, but most are just discreet about it.

> To ensure programmers will use good style, Bentley will lack goto. To
> break out of nested loops, you can use the breakout statement.

This worries me.  When I need to implement a finite state autonomon I
usually use goto.  For that purpose it is by far the clearest and least
error prone method C offers.  How do I do this in Bentley?  Infinite
loop, state variable and breakout?  I would argue that that's less
clear and unless your compiler is very clever it will generate worse
assembly in this (common) case.

> Finally, there will be two modes: hosted and standalone. The
> standalone keyword changes this. Hosted mode can access print to
> stdout and stderr, read from stdin, new, renew (like realloc), delete,
> and a string type.

I don't understand.
--
John Stalker
School of Mathematics
Trinity College Dublin
tel +353 1 896 1983
fax +353 1 896 2282

Re: [9fans] A new language for Plan 9

[Bakul Shah]

On Fri, 02 May 2008 08:49:24 BST John Stalker <stalker@maths.tcd.ie>  wrote:
> > To ensure programmers will use good style, Bentley will lack goto. To
> > break out of nested loops, you can use the breakout statement.
>
> This worries me.  When I need to implement a finite state autonomon I
> usually use goto.  For that purpose it is by far the clearest and least
> error prone method C offers.  How do I do this in Bentley?  Infinite
> loop, state variable and breakout?  I would argue that that's less
> clear and unless your compiler is very clever it will generate worse
> assembly in this (common) case.

If he provides proper tail-recursion or even a switch
statment you can implement FSAs quite easily.

Clearly he has a lot of enthusiasm but I don't understand why
he is squandering it on implementing boring old language
ideas.  Why not a language for mashing tree structured data
or graphs or for parallel programming or modelling physics of
real objects or something.  He'd learn even if he fails.

Hoare's 1973 paper "Hints on Programming Language Design"
might be relevant to what he is attempting (look for
STAN-CS-73-403)

Re: [9fans] A new language for Plan 9

[Martin Neubauer]

* Bakul Shah (bakul+plan9@bitblocks.com) wrote:
> Clearly he has a lot of enthusiasm but I don't understand why
> he is squandering it on implementing boring old language
> ideas.

Old ideas have the benefit of already being there.