Lex/YACC Question

theNbomr · 03-20-2010, 07:26 PM

But English sucks for stuff like that. That's why things like BNF were invented. And once you've created that unambiguous description, you're pretty much just a 'make' away from the final product.
--- rod.

MTK358 · 03-20-2010, 07:28 PM

EDIT: Be sure to note the difference between Statement and Statements.

Statements:
(any amount of Statement's)

Statement:
Expression \n
---or---
while Expression \n Statements loop \n
---or---
if Expression \n Statements fi \n
---or---
if Expression \n Statements else \n Statements fi \n

Expression:
(any expression that evaluates to a value, including math ops, variable assignments , boolean ops, function calls, etc.)

Sergei Steshenko · 03-20-2010, 07:30 PM

Quote:

Originally Posted by theNbomr

But English sucks for stuff like that. That's why things like BNF were invented. And once you've created that unambiguous description, you're pretty much just a 'make' away from the final product.
--- rod.

Of course. But the sarcastic point is that BNF itself is explained in (terms of) English.

So IMO it would make sense for MTK358 to first explain what he wants in English, and then his explanation/desires will be formalized into something conceptually similar to BNF and friends.

MTK358 · 03-20-2010, 07:33 PM

Quote:

Originally Posted by Sergei Steshenko

So IMO it would make sense for MTK358 to first explain what he wants in English

I already did that in my previous post.

Sergei Steshenko · 03-20-2010, 07:38 PM

Quote:

Originally Posted by MTK358

Statements:
(any amount of Statements)

Statement:
expression \n
---or---
while Expression \n Statements loop \n
---or---
if Expression \n Statements fi \n
---or---
if Expression \n Statements else \n Statements fi \n

Expression:
(any expression that evaluates to a value, including math ops, variable assignments , boolean ops, function calls, etc.)

I suggested first English, not that formal explanation.

For example:

do you want variables to be declared before they are used ?
do you want to have a notion of assignment in its C/Perl sense ?
do you want an assignment to return a value, thus allowing nested assignments and "dirty" "C" code like 'if(a = b)' (yes, just one '=')
do you want to have the possibility for while/if/else body to have more than one "action" (by "action" I mean C/Perl executable statement separated by ';') ?

MTK358 · 03-20-2010, 07:49 PM

Quote:

Originally Posted by Sergei Steshenko

do you want variables to be declared before they are used ?

For the interpreter version, I don't care, but for the compiler version, yes.

Quote:

Originally Posted by Sergei Steshenko

do you want to have a notion of assignment in its C/Perl sense ?

You mean assigning an expression to a variable? Yes, using the := operator. +=, -=, *=, /=, and %= (and maybe more) will also be available.

Quote:

Originally Posted by Sergei Steshenko

do you want an assignment to return a value, thus allowing nested assignments and "dirty" "C" code like 'if(a = b)' (yes, just one '=')

Yes.

But := will be the assignment op (like "=" in C) (it will evaluate to the new value of the variable being assigned to), and = will be the comparison op (like "==" in C).

Quote:

Originally Posted by Sergei Steshenko

do you want to have the possibility for while/if/else body to have more than one "action" (by "action" I mean C/Perl executable statement separated by ';') ?

What do you mean?
They can contain multiple statements, each on its own line (no semicolons). See my "pseudo-BNF" description that I previously posted.

Sergei Steshenko · 03-20-2010, 07:57 PM

Quote:

Originally Posted by MTK358

...

What do you mean?
They can contain multiple statements, each on its own line (no semicolons).

In "C"

Code:

if(a > b)
  c = d;
  e = f; // misleading indentation !!!

the 'e = f' will always be executed - regardless of 'a > b'.

So, in C/Perl (and not only them) there is a notion of code block (denoted by '{', '}'), i.e. if the above both assignments are both meant to be conditional, one should write

Code:

if(a > b)
  {
  c = d;
  e = f;
  }

.

So, do you want code blocks ? And, I guess, you've already guessed that I'm pushing you towards 'code block' language construct.

MTK358 · 03-20-2010, 08:01 PM

In my language they will all be blocks by default, there will be no way not to make a block.

For example:

Code:

if a > b
    c = d # Both are executed only if (a > b) != 0
    e = f
fi

smeezekitty · 03-20-2010, 08:03 PM

Quote:

Originally Posted by MTK358

In my language they will all be blocks by default, there will be no way not to make a block.

For example:

Code:

if a > b
    c = d # Both are executed only if (a > b) != 0
    e = f
fi

Bash style?

MTK358 · 03-20-2010, 08:05 PM

Quote:

Originally Posted by smeezekitty

Bash style?

Kind of.

I was thinking something like bash/BASIC-like syntax with C-like functionality.

smeezekitty · 03-20-2010, 08:07 PM

Quote:

Originally Posted by MTK358

Kind of.

I was thinking something like bash/BASIC syntax with C capabilites.

You mean like pascal?

http://en.wikipedia.org/wiki/Pascal_...ng_language%29

MTK358 · 03-20-2010, 08:09 PM

I guess the control structures look kind of like what I had in mind, but the rest looks quite different.

Sergei Steshenko · 03-20-2010, 08:20 PM

Quote:

Originally Posted by MTK358

In my language they will all be blocks by default, there will be no way not to make a block.

For example:

Code:

if a > b
    c = d # Both are executed only if (a > b) != 0
    e = f
fi

OK, so you want blocks, but the question remains - if you have an 'if' with and 'else' how do you denote the 'if' block and how do you denote the 'else' block ?

MTK358 · 03-21-2010, 06:25 AM

The keyword if followed by an expression and a newline starts a block.

The keyword fi ends a block started by if

The keyword else on its own line can optionally be inserted within the block, separating it.

I am really getting annoyed by this. All this is explained clearly and simply in post #62. If you can't understand that diagram, then of all seriousness I don't think you're the right one to help me with this.

And BNF is described in English not because English is good, but because it's the only thing you can be sure a person understands beforehand.

MTK358 · 03-21-2010, 08:29 AM

I wrote a BNF grammar for most of the language:

EDIT: Revised

Everything in []s is optional
Everything in {}s can be repeated 0 or more times.

Code:

<stmts> ::= |
            { <stmt> }

<stmt> ::= <expr> "\n" |
           "if" <expr> "\n" <stmts> [ "else" "\n" <stmts> ] "fi" \n |
           "while" <expr> "\n" <stmts> "loop" "\n"

<expr> ::= <int> |
           <float> |
           <string> |
           <id> |
           <id> "++" |
           <id> "--" |
           "++" <id> |
           "--" <id> |
           <id> ":=" <expr> |
           <expr> "+" <expr> |
           <expr> "-" <expr> |
           <expr> "*" <expr> |
           <expr> "/" <expr> |
           <expr> "%" <expr> |
           <expr> "=" <expr> |
           <expr> "!=" <expr> |
           <expr> ">" <expr> |
           <expr> "<" <expr> |
           <expr> ">=" <expr> |
           <expr> "<=" <expr> |
           "not" <expr> |
           <expr> "and" <expr> |
           <expr> "nand" <expr> |
           <expr> "or" <expr> |
           <expr> "nor" <expr> |
           <expr> "xor" <expr> |
           <expr> "xnor" <expr> |
           "(" <expr> ")" |
           <id> "(" [ <expr> ] { "," <expr> } ")"