Implement the named labels for assembly

Question

Let's assume we've got an imaginary assembler. The assembler supports numerical labels. An infinite loop example:

:1
    jmp 1

Your task is, to write a preprocessor for this assembler supporting named labels (instead of numerical ones; with up to 8 characters in length), so the label name is replaced to it's numerical value.

The assembler uses a dot (.) to express a character constant, for example:

mov dl, .A

Will store ASCII(65) - 'A', in the register dl. The assembler supports string constants in double quotes too:

:20
db "Hello, world!"

So, to sum things up, you need to be aware of character constants, and the fact that you can't perform any substitution inside quote-enclosed strings.

The label declaration character and reference character is up to your choice, but for the sake of completness I will use a $ to refer to a label, and @ to denote a label.

The label' number might be any natural number not equal to zero.

You may assume that assembly on the input is perfectly valid.

Let's look at a few examples:

@loop
    jmp $loop

:1
    jmp 1

---

jmp $skip
mov dl, .@
@skip
cmp dl, .$

jmp 1
mov dl, .@
:1
cmp dl, .$

---

@x
db "$a @b"
@a
    jmp $a
    @b
        jmp $b

:1
db "$a @b"
:2
    jmp 2
    :3
        jmp 3

This is a code golf, so the smallest answer wins. Standard rules and loopholes apply.

You can safetly assume that the user won't use numerical labels anymore, and the labels contain just A-Za-z letters.

Label declaration and reference character has to be different.

Answer 1

JavaScript (ES6), 96 bytes

Uses @label for a declaration and %label for a reference.

s=>s.replace(o=/^\s*@(.+)/gm,(_,s)=>':'+(o[s]=++n),n=0).replace(/".*?"|%(\w+)/g,(s,x)=>x?o[x]:s)

Try it online!

Commented

(alternate slash characters are used below to prevent the SE syntax highligter from going mad)

s =>                  // s = input string
  s.replace(          // 1st pass:
    o =               //   assign the regex object to o; we'll use it to store the labels
    ∕^\s*@(.+)∕gm,    //   look for [start] + [optional whitespace] + '@' + [label]
    (_, s) =>         //   replace with:
      ':' +           //     ':' + label ID
      (o[s] = ++n),   //     increment n and store the label in o
    n = 0             //   start with n = 0
  )                   // end of replace()
  .replace(           // 2nd pass:
    ∕".*?"|%(\w+)∕g,  //   look for either ".*?" (non-greedily) or %label
    (s, x) =>         //   replace with:
      x ? o[x]        //     o[label] if x is defined,
        : s           //     or the original string otherwise
  )                   // end of replace()

Answer 2

PHP (7.4), 148 143 bytes

-5 bytes by stealing RegEx idea from @Arnauld.

<?=preg_replace_callback('/".*?"|([@_]\w+)/',fn($m)=>$m[1]?[':'][$m[0][0]>A].($_GET[$w=substr($m[0],1)]?:$_GET[$w]=++$_GET[0]):$m[0],$argv[1]);

Try it online!

Define labels with @ and refer to them with _.

Captures any @<label> or _<label> which aren't inside double quotes ("...") and replaces them with a unique number for every unique label, starting from 1. Also adds the : when replacing labels that have @ before them.

PHP's global variable $_GET is used to store and access last used id and unique ids for each label inside the arrow function. $_GET[0] holds last used id and $_GET[<label>] holds unique id for <label>.

Answer 3

Retina 0.8.2, 97 bytes

ms(1`^@
:@
+`(.*^:(1*).*?^)@
$1:1$2@
^:1*
:$.&
\$(.+?)$(?<=(?=.*^:(\d+)@\1$).+)
$2
^(:\d+).*?$
$1

Try it online! Uses the provided symbols (could save 1 byte by using a different symbol). Explanation:

ms(

Run everything in single multiline mode, where . matches newlines and ^ and $ match at the beginning and end of each line.

1`^@
:@

Prefix a : to the first label.

+`(.*^:(1*).*?^)@
$1:1$2@

Number the labels in unary.

^:1*
:$.&

Convert to decimal.

\$(.+?)$(?<=(?=.*^:(\d+)@\1$).+)
$2

Find all matching references and replace them with the number.

^(:\d+).*?$
$1

Delete the label names.

Answer 4

Python 3, 136 117 bytes

New and improved:

import re;d={}
f=lambda s:re.sub(r'(".+?")|([@$])(\w+)',lambda m:m[1]or d.setdefault(m[3],f":{m.end()}")[m[2]<='@':],s)

Try it online!

Uses the position of the match.end() as the integer for a label.

Old code

import re;d={};n=[*range(9999)]
f=lambda s:re.sub(r'(".+?")|([@$])(\w+)',lambda m:m[1]or d.setdefault(m[3],f":{n.pop()}")[m[2]<='@':],s)

Try it online!

Labels start at 9998 and count downward. Some numbers may be skipped. It will handle about 5000 labels, which should be enough for any sane asm program, but the number can be increased at 1 byte for each order of magnitude increase in range.

Answer 5

Perl 5 (-p), 57 bytes

s/^\s*\K@(\w+)/":".($$1=++$,)/mge;s/".*?"\K|\$(\w+)/$$1/g

Try it online!

• $, : global variable used as counter
• s/^\s*\K@(\w+)/":".($$1=++$,)/mge : replaces labels "@[label]" by ":[counter]", increment counter and store counter in variable label name's variable
• ".*?"\K| : regex part used to ignore strings between ""
• s/...\$(\w+)/$$1/g : replaces references by variables value

Answer 6

Javascript ES6, 71 char

s=>s.replace(/(;.*|".*?")|:(:?)(\w+)/gm,(m,s,c,l)=>s||c+parseInt(l,36))

Conditions:

• :: to introduce the label
• : to use the label
• Labels are case insensitive
• ; starts comment till the end of the line
• Quoted strings may not have " or linebreak inside
• Charecter constants may not be followed by colon or letter like .:a or .::
• Classic numeric labels starting from : are disallowed

Test:

console.log((

s=>s.replace(/(;.*|".*?")|:(:?)(\w+)/gm,(m,s,c,l)=>s||c+parseInt(l,36))

)(`
::loop
    jmp :loop

jmp :skip
mov dl, .:
::skip
cmp dl, .:

::x
db ":a ::b"
::a
    jmp :a
    ::b
        jmp :b
`))

.as-console-wrapper.as-console-wrapper{max-height:100vh}