Re: neos Universal Compiler and FORTRAN IV

On 3/29/2025 1:01 PM, Mr Flibble wrote:

On Sat, 29 Mar 2025 11:53:33 -0600, Jeff Barnett wrote:
 

On 3/29/2025 9:46 AM, Mr Flibble wrote:

On Sat, 29 Mar 2025 00:22:08 -0600, Jeff Barnett wrote:
>

On 3/28/2025 1:27 PM, Mr Flibble wrote:

On Fri, 28 Mar 2025 13:05:10 -0600, Jeff Barnett wrote:
>

On 3/28/2025 8:14 AM, Mr Flibble wrote:

On Fri, 28 Mar 2025 14:01:40 +0200, Mikko wrote:
>

On 2025-03-19 16:11:18 +0000, Mikko said:
>

On 2025-03-19 14:32:30 +0000, Mr Flibble said:
>

On Wed, 19 Mar 2025 12:26:20 +0200, Mikko wrote:
>

On 2025-03-18 14:08:50 +0000, Mr Flibble said:
>

On Tue, 18 Mar 2025 15:59:45 +0200, Mikko wrote:
>

On 2025-03-17 16:53:01 +0000, Mr Flibble said:
>

On Mon, 17 Mar 2025 11:21:05 +0200, Mikko wrote:
>

On 2025-03-16 18:40:42 +0000, Mr Flibble said:
>

On Sun, 16 Mar 2025 12:28:24 +0200, Mikko wrote:
>

On 2025-03-15 15:08:47 +0000, Mr Flibble said:
>

On Sat, 15 Mar 2025 11:55:52 +0200, Mikko wrote:
>

On 2025-03-15 04:00:52 +0000, Mr Flibble said:
>

Hi!
>
neos universal compiler (that can compile any
programming language)
is successfully running the tokenization stage
tokenizing a program written in the neos reference
language. #cpp #coding #compiler #compsci #gamedev
>
/Flibble

>
Can it tokenize FORTRAN 60 or FORTRAN IV ?

>
ANY programming language.
>
/Flibble

>
How is neos configured to tokenize FORTRAN 60 ?

>
The same way you would configure it for any other
programming language.

>
If it is configured the same for each programming language
then how does it know how to tokenize?

>
You configure it by providing a language specific neosBNF
schema (grammar)
file (an input to the compilalation process).

>
Is there a neosBNF schema that describes the tokens of
FORtRAN 66 or Algol 60?

>
Not yet.

>
The definition of string literal of Algol 60 would be a good
example of something that cannot be defined with a regular
expression and is therefore impossible or at least complicated
with an ordinary tokenizer.

>
neos does not use regular expressions and the neos grammar is
context sensitive (i.e. not context free).

>
A context free grammar should be enough for tokenizing strings of
Algol 60.
>
Tokenizing statements of FORTRAN 66 is a harder problem.

>
Some test cases for a FORTRAN IV tokenizer:
>
            REALITY IS, AS I MAY LOGICALLY SAY, REAL LOGICAL = WHAT
            I SAY
>
Note that these lines begin with 8 spaces.
>
Tokens on the first line:
keyword REAL identifier ITYIS comma identifier ASIMAY end of
statement
>
Tokens on the second line:
keyword LOGICAL identifier LYSAY comma identifier REAL end of
statement
>
Tokens on the third line:
identifier LOGICAL assignment symbol identifier WHAT I SAY
>
Parser needs to know whether an instance of REAL or LOGICAL is an
identifier or keyword.
>
I don't understand how neos can be configred to tokenize the above
test lines and other FORTRAN IV statements.

>
If a native FORTRAN IV compiler can parse it then neos can parse it
as long as it is somehow expressable in the neos attribute grammar.

>
But that was the question: Can it be parsed in a neos expressible
grammar? Note that those early Fortran grammars were 1) not finite
state, 2) not context free, and 3) inherently ambiguous.It was
possible,
for example, to write a "format" and an "assignment" statement that
were character to character identical. As an aside, a friend who
wrote an early Fortran compiler that compiled such a chimera so that
it would do an assignment if executed and could be referenced as a
format from an IO statement.

>
The tokenisation step is entirely optional in neos which instead
deploys a configurable multi-stage parsing "pipeline" so again there
should be no problem if FORTRAN IV is expressable in the neos
attribute grammar given the problem is itself not intractable as
FORTRAN IV compilers already exist.

>
I REPEAT -- But that was the question: Can it be parsed in a neos
expressible grammar?

>
There should be no problem if FORTRAN IV is expressable in the neos
attribute grammar given the problem is itself not intractable as
FORTRAN IV compilers already exist.

That reasoning for your answer has two possible explanations: 1) You
think it's cute and 2) You are saying, by implication, that the neos
attribute grammar is Turing complete (though an implementation is finite
restricted).
>
If it is 1, continue to enjoy yourself. If it is 2, do you have a proof
sketch? If true, an example of what your attribute grammar should be
able to parse into tokens is to take a string of digits (0 to 9) and
break those digit strings into prime numbers if possible and issue an
error message if it is impossible.
>
I'm really curious about just what you are claiming.

 There is no reason to break those digit strings into prime numbers at the
tokenisation stage as tokenisation is entirely optional in neos.  The neos
compilation framework is indeed Turing Complete.

Your style of interacting is developing before my eyes. I know who your guru is: Has Olcott started calling you grasshopper? I didn't know he had a dialectic disciple. Congratulations to both of you.
--
Jeff Barnett