meta Meta language released

[Kaj de Vos]

Get ready for new Atari users! ?

http://rebolforum.com/index.cgi?f=printtopic&permalink=Kaj de Vos28-Oct-2021/13:19:39-7:00&archiveflag=new

[ilmenit]

8 minutes ago, Kaj de Vos said:

Ah, you can't iterate a constant. You first need to assign flags to a variable reference.

how to do it? None of these do the job:

flags-ptr: flags 

flags-ptr: block! flags

flags-ptr: [flags]

[Kaj de Vos]

The first should work. What is the error message?

[ilmenit]

For the code

    flags= make static binary! sizepl
    flags-ptr: flags

    for-all flags-ptr 
    [ 
;        change flags-ptr 1 ; "change" has no value
        poke flags-ptr 0 1
    ]

I receive

** Script error: "for-all" does not work on types combination:
Parameter types:
1: static-binary! [flags-ptr
    [
     ...
2: expressions list block! [
    [
        poke f...
At: [
    for-all flags-ptr
    [
        poke flags-ptr 0 1
...
In: [
    count: 0
    i: 0
    flags= make static binary! si...
Near: [
    while iter <= 10
    [
        count: 0
        i: ...

 

[Kaj de Vos]

Thanks, made another fix.

[Kaj de Vos]

9 minutes ago, ilmenit said:

poke flags-ptr 0 1

Please don't call it a pointer. It's a high-level series.

 

Your indexing is off by one. Indexing in REBOL and Meta is 1-based. In the future, this will produce a bounds error.

[ilmenit]

4 minutes ago, Kaj de Vos said:

Thanks, made another fix.

now the compilations is getting error later on pick:

** Script error: cannot use pick on none! value
** Where: has-arguments any do if case either compile-expression foreach unless either compile-expression foreach unless either compile-expression foreach emit-C-program either do either either either -apply-
** Near: has-arguments third arguments-of code all [
    constant: se...

the code:

natural! [count i]

while iter <= 10 
[
    count: 0
    i: 0
    flags= make static binary! sizepl
    flags-ptr: flags

    for-all flags-ptr 
    [ 
;        change flags-ptr 1 ; "change" has no value
        poke flags-ptr 0 1
    ]

    i: 0
    while i <= size
    [
        p: pick flags i
        if p = 1 
        [
            prime: ( i * 2 ) + 3
            k: prime + i 
            while k <= size
            [
                poke flags k 0
                k: k + prime
            ]
            count: count + 1
        ]
        i: i + 1
    ]
    iter: iter + 1
]

 

[Kaj de Vos]

It's not on your pick, it's in REBOL code. Bug in the compiler.

[Kaj de Vos]

Should be fixed now.

[ilmenit]

24 minutes ago, Kaj de Vos said:

Should be fixed now.

Compiling file sieve.meta

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  1248  100   344  100   904    246    647  0:00:01  0:00:01 --:--:--   895

** Script error: insert does not allow none! for its series argument
** Where: either do if case either compile-expression foreach unless either compile-expression foreach unless either compile-expression foreach emit-C-program either do either either either -apply-
** Near: either any [
    not has-arguments second arguments-of code

[Kaj de Vos]

16 hours ago, ilmenit said:

if you see there any places to improve the code, would be great to get your hints here.

Here goes.

16 hours ago, ilmenit said:

iter: 1

iter is inferred as integer!. You could optimise it like this:

byte! iter

It wouldn't matter much for speed here, but it would make the executable a bit smaller. But you don't need iter, because you can just do:

repeat 10 [

 

16 hours ago, ilmenit said:

unsafe!!! constant reference volatile byte! [     ; OS     RTCLOK2= ~13     RTCLOK3= ~14 ]

Address assignments increment automatically, so you can write:

unsafe!!! constant reference volatile byte! [
    ; OS
    RTCLOK2= ~13
    RTCLOK3
]

 

16 hours ago, ilmenit said:

RTCLOK2: 0 RTCLOK3: 0

RTCLOK3: RTCLOK2: 0

is shorter, in source and possibly also in binary code, depending on the back-end.

 

The first

16 hours ago, ilmenit said:

i: 0

is superfluous.

 

You can fill the array with

change/repeat flags 1 sizepl

 

The second

16 hours ago, ilmenit said:

i: 0

is off by one.

 

16 hours ago, ilmenit said:

while i <= size

for starts with one by default to match series indexes. Just do

for i size [

 

16 hours ago, ilmenit said:

p: pick flags i         if p = 1

p is inferred as integer!. This is inefficient. But you don't need p:

if (pick flags i) = 1

Experienced REBOL programmers like to simplify this:

if 1 = pick flags i [

Unlike REBOL, Meta supports the C style of treating 0 as an option value, so you can just do

if pick flags i [

 

16 hours ago, ilmenit said:

( i * 2 ) + 3

This can be optimised as

i << 1 + 3

But watch out that you need to adjust for 1-indexing.

Whether this produces better code depends on the back-end.

 

16 hours ago, ilmenit said:

count: count + 1

The remaining increment can be optimised as

increment count

 

[ilmenit]

17 minutes ago, Kaj de Vos said:

You can fill the array with

change/repeat flags 1 sizepl

    flags= make static binary! sizepl    
    change/repeat flags 1 sizepl  

 

** Script error: "change/repeat" does not work on types combination:
Parameter types:
1: static-binary! [flags 1 sizepl
    fo...
2: byte! [1 sizepl
    for i si...
3: natural16! [sizepl
    for i size...
At: [
    change/repeat flags 1 sizepl
    for i size
    [
 ...
In: [
    count: 0
    flags= make static binary! sizepl
    ...
Near: [
    repeat 10
    [
        count: 0
        flags= mak...

 

Quote

Your indexing is off by one. Indexing in REBOL and Meta is 1-based. In the future, this will produce a bounds error.

It's a topic for a flame war, but are you sure it's good idea to follow 1-based indexing of Rebol? You are dropping compatbility with Rebol anyway.

Many algorithms working on arrays (esp. of more than 1 dimension) are much simpler with 0-based indexing (and if you are targeting 8bit machines then you will work with them primarily). Modern languages that primarily 1-based have ability to set arbitrary indices (Julia, Pascal). Also nowadays "a must" are bindings to different libraries... and most of them are done in 0-based languages.

Quote

 

This can be optimised as

i << 1 + 3

But watch out that you need to adjust for 1-indexing.

 

This one of examples why 1-based indexing is not that great idea.

 

[Kaj de Vos]

1 hour ago, ilmenit said:

Compiling file sieve.meta

  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100  1248  100   344  100   904    246    647  0:00:01  0:00:01 --:--:--   895

** Script error: insert does not allow none! for its series argument
** Where: either do if case either compile-expression foreach unless either compile-expression foreach unless either compile-expression foreach emit-C-program either do either either either -apply-
** Near: either any [
    not has-arguments second arguments-of code

Made a partial fix.

 

Out of time for today.

[zbyti]

Quote

Address assignments increment automatically

won't work for a last assignment in block:

 

** Script error: "=" does not work on types combination:
Parameter types:
1: unset word! [VCOUNT]

 

Edited November 19, 2021 by zbyti

[Kaj de Vos]

8 minutes ago, ilmenit said:

"change/repeat" does not work on types combination

Made a fix.

[Kaj de Vos]

9 minutes ago, zbyti said:

won't work for a last assignment in block:

 

** Script error: "=" does not work on types combination:
Parameter types:
1: unset word! [VCOUNT]

 

That's a comparison, you should not have a " = " in there.

[Kaj de Vos]

12 minutes ago, ilmenit said:

It's a topic for a flame war, but are you sure it's good idea to follow 1-based indexing of Rebol?

Yes, and yes, it's more human-friendly. We have plenty of unfriendly languages already.

13 minutes ago, ilmenit said:

Also nowadays "a must" are bindings to different libraries... and most of them are done in 0-based languages.

Doesn't matter, you just compensate.

14 minutes ago, ilmenit said:

This one of examples why 1-based indexing is not that great idea.

It's still math. Surely, a mathematically inclined person can easily adapt.

[zbyti]

23 minutes ago, Kaj de Vos said:

That's a comparison, you should not have a " = " in there.

No.

 

unsafe!!! constant reference volatile byte! [
  RTCLOK=  ~14
  P0PF=    ~2C0
  HPOSM0=  ~D004
  GRAFM=   ~D011
  COLPM0
  COLPM1
  COLBK=   ~D01A
  RANDOM=  ~D20A
  WSYNC=   ~D40A
  VCOUNT
]

this don't work, last assignment must have explicit address.

[TGB1718]

46 minutes ago, Kaj de Vos said:

1 hour ago, ilmenit said:

Also nowadays "a must" are bindings to different libraries... and most of them are done in 0-based languages.

Doesn't matter, you just compensate.

Why not give the programmer the option to set the base like some languages already do

Best of both worlds

 

Atari ST BASIC uses OPTION BASE  which can be 0 or 1

Edited November 19, 2021 by TGB1718
Update

[Kaj de Vos]

32 minutes ago, zbyti said:

No.

 

unsafe!!! constant reference volatile byte! [
  RTCLOK=  ~14
  P0PF=    ~2C0
  HPOSM0=  ~D004
  GRAFM=   ~D011
  COLPM0
  COLPM1
  COLBK=   ~D01A
  RANDOM=  ~D20A
  WSYNC=   ~D40A
  VCOUNT
]

this don't work, last assignment must have explicit address.

I see, that problem takes more work.

[Kaj de Vos]

11 minutes ago, TGB1718 said:

Why not give the programmer the option to set the base like some languages already do

Best of both worlds

 

Atari ST BASIC uses OPTION BASE  which can be 0 or 1

Or worst of both worlds. It would take a lot of double implementations. Better standardise.

[sanny]

5 minutes ago, Kaj de Vos said:

Better standardise.

But not to 1 IMHO. Feels unintuitive for normal programmers ?

 

Just my 2 cents....

[Kaj de Vos]

https://language.metaproject.frl/#for

 

Normal programmers are not the target audience.

[TGB1718]

10 minutes ago, sanny said:

But not to 1 IMHO. Feels unintuitive for normal programmers ?

I agree, 1 just seems wrong and causes it's own problems

[Kaj de Vos]

Agree to disagree then. Both have issues. Meta chooses for the human-friendly way.