update ts-minifier

ts-minifier.py

@@ -1,29 +1,66 @@
 # Copyright (c) 2021 bleck9999                                                                                                                
 # https://github.com/bleck9999/ts-minifier                                                                                                    
-# Version: 91345ce1
+# Version: 700ce8d
 
 import argparse
-import re
+import itertools
+from string import ascii_letters
 
-# if is not included because it's already 2 characters
+auto_replace = False
-sub_funcs = {'while': "_h", 'print': "_p", 'println': "_l", 'mountsys': "_s", 'mountemu': "_e", 'readsave': "_r",
+stdlib = ['if', 'while', 'print', 'println', 'mountsys', 'mountemu', 'readsave', 'exit', 'break', 'dict', 'setpixel',
-             'exit': "_q", 'break': "_b", 'dict': "_d", 'setpixel': "_y", 'readdir': "_i", 'copyfile': "_c",
+          'readdir', 'copyfile', 'mkdir', 'ncatype', 'pause', 'color', 'menu', 'emu', 'clear', 'timer', 'deldir',
-             'mkdir': "_k", 'ncatype': "_n", 'pause': "_w", 'color': "_a", 'menu': "__", 'emu': "_u",
+          'fsexists', 'delfile', 'copydir', 'movefile', 'payload', 'readfile', 'writefile', 'setpixels', 'printpos',
-             'clear': "_x", 'timer': "_t", 'deldir': "_g", 'fsexists': "_f", 'delfile': "_z", "copydir": "c_",
+          'emmcread', 'emmcwrite', 'emummcread', 'emummcwrite', 'escapepath', 'combinepath', 'cwd', 'power',
-             "movefile": "_v", "payload": "_j", "readfile": "_o", "writefile": "_W", "setpixels": "_Y", "printpos": "_P",
+          'fuse_patched', 'fuse_hwtype']
-             "emmcread": "_E", "emmcwrite": "_F", "emummcread": "_R", "emummcwrite": "_S", "escapepath": "_X",
-             "combinepath": "_A", "cwd": "_D", "power": "_O", "fuse_patched": "_M", "fuse_hwtype": "_N"}
-replace_functions = False
 
 
-def wantsumspace(s: str):
+class Code:
-    for c in s.lower():
+    def __init__(self, strings, comments, script):
-        if (ord(c) < 97 or ord(c) > 122) and (ord(c) != 95) and not (c.isnumeric()):
+        counter = 0
+        strings_comments = sorted(strings + comments)
+        bounds = [0] if strings_comments[0][0] != 0 else []
+        for val in strings_comments:
+            if counter and (bounds[counter - 1] == val[0]):
+                bounds[counter - 1] = val[1]
+            else:
+                bounds += [val[0], val[1]]
+                counter += 2
+        bounds.append(len(script))
+        code = []
+        i = 2 if len(bounds) % 2 else 1
+        while i < len(bounds):
+            code.append((bounds[i - 1], bounds[i], script[bounds[i - 1]:bounds[i]]))
+            i += 2
+        self.sections = sorted(strings_comments + code)
+        self.strings = strings
+        self.comments = comments
+        self.code = code
+        # self.string_comments = strings_comments
+        self.rawcode = "".join([x[2] for x in sorted(self.code+self.strings)])
+
+    def getafter(self, ch: int):
+        ch += self.comments[-1][1] if self.comments else 0
+        for strcom in self.strings:
+            if strcom[0] >= ch:
+                return strcom
+        return None
+
+    def nextch(self, ch: int, reverse: bool):
+        rawcontent = self.rawcode
+        if ((ch+1 >= len(rawcontent)) and not reverse) or \
+                ((ch-1 < 0) and reverse):
+            return ''
+        return rawcontent[ch-1] if reverse else rawcontent[ch+1]
+
+
+def isidentifier(s: str):
+    for c in s:
+        if c not in (ascii_letters + '_'):
             return False
     return True
 
 
-def commentstartswhere(s: str):
+def hascomment(s: str):
     quoted = False
     for c in range(len(s)):
         if s[c] == '"':
@@ -33,29 +70,233 @@ def commentstartswhere(s: str):
     return None
 
 
-def minify(script: str):
+def parser(script: str):
-    # currently ts does not seem to allow 's to mark a quote
+    comments = []  # [(start, end, content)]
-    # (https://github.com/suchmememanyskill/TegraExplorer/blob/tsv3/source/script/parser.c#L173)
+    strings = []
-    # im fine with that, it makes doing this a lot easier
+    commented = False
-    # strings = script.split(sep='"')
+    quoted = False
+    strstart = -1
+    commentstart = -1
+    for c in range(len(script)):
+        if script[c] == '#' and not quoted:
+            commented = True
+            commentstart = c
+        elif (script[c] == '\n' and not quoted) and commented:
+            comments.append((commentstart, c + 1, script[commentstart:c + 1]))
+            commented = False
+        elif script[c] == '"' and not commented:
+            if not quoted:
+                strstart = c
+                quoted = True
+            else:
+                strings.append((strstart, c + 1, script[strstart:c + 1]))
+                quoted = False
+
+    script = Code(strings, comments, script)
+
+    # guess i should do a breakdown of step 3
+    # we need to be able to read:
+    #    variable creation | a = 15, array.foreach("a")
+    #    defining a function | funcname = {function body}
+    #    calling a function | funcname(arguments) for stdlib functions, funcname(<optional> any valid ts) for user defined
+    #    member calling | object.member(possible args)
+    #        we don't need to check if it's valid syntax or not so we dont need to know the type of object that's nice
+    #        this can actually be chained which is pretty annoying
+    #    operators? i dont think it actually matters to us
+    #    *statements* | a
+    #        these can be delimited by anything that isn't a valid identifier.
+    #        fuck me clockwise
+    #
+    # other notes:
+    #   we minify the script a little before parsing, so there is no unnecessary whitespace or comments
+    #   we are assuming the input script is valid syntax
+
+    userobjects = {}
+    usages = {}
+    hexxed = False
+    ismember = False
+    quoted = False
+    strscript = script.rawcode
+    start = len(strscript) + 1
+    for ch in range(len(strscript)):
+        if (strscript[ch-1] == '0' and strscript[ch] == 'x') and not quoted:
+            hexxed = True
+        elif isidentifier(strscript[ch]) and not (hexxed or quoted):
+            if start > ch:
+                start = ch
+            else:
+                pass
+        elif hexxed and strscript[ch].upper() not in "0123456789ABCDEF":
+            hexxed = False
+        elif strscript[ch] == '"':
+            quoted = not quoted
+        elif not quoted:
+            if start != len(strscript)+1:  # if we actually had an identifier before this char
+                identifier = strscript[start:ch]
+                if identifier in usages:
+                    usages[identifier].append(start)
+                elif strscript[ch] == '=' and strscript[ch+1] != '=':
+                    isfunc = script.nextch(ch, False) == '{'
+                    userobjects[identifier] = "func" if isfunc else "var"
+                    usages[identifier] = [start]  # declaration is a usage because i cant be arsed
+                elif not ismember:  # not an assignment (or member) but also haven't seen this name before
+                    usages[identifier] = [start]
+                    # fuck it we are using a fucking list of fucking stdlib functions i just fucking cant im adding tsv3
+                    # to the fucking esolangs wiki have a good day
+                    if identifier not in stdlib:
+                        userobjects[identifier] = "var"
+            if strscript[ch] == '.':
+                if ismember:  # we check if there's a . after a ), if there is we know that there's nothing to do here
+                    continue
+                ismember = True
+                # we don't really care about anything else
+            elif strscript[ch] == '(':
+                if ismember:
+                    if "foreach" == strscript[start:ch]:  # array.foreach takes a variable name as an arg (blame meme)
+                        name = script.getafter(ch)[2].replace('"', '')
+                        if name in userobjects:
+                            usages[name].append(start)
+                        else:
+                            usages[name] = []
+                            userobjects[name] = "var"
+                    else:
+                        pass
+            elif strscript[ch] == ')':
+                ismember = script.nextch(ch, False) == '.'
+            start = len(strscript) + 1
+
+    return minify(script, userobjects, usages)
+
+
+def minify(script: Code, userobjects, usages):
+    # the space saved by an alias is the amount of characters currently used by calling the function (uses*len(func))
+    # minus the amount of characters it would take to define an alias (len(alias)+len(func)+2), with the 2 being the
+    # equals and the whitespace needed for a definition
+    # obviously for a rename you're already defining it so it's just the difference between lengths multiplied by uses
+    short_idents = [x for x in (ascii_letters+'_')] + [x[0]+x[1] for x in itertools.permutations(ascii_letters+'_', 2)]
+    short_idents.pop(short_idents.index("if"))
+    mcode = script.rawcode
+    aliases = []
+    for uo in [x for x in userobjects]:
+        if userobjects[uo] not in ["var", "func"]:
+            continue
+        tmpcode = ""
+        otype = userobjects[uo]
+        uses = len(usages[uo])
+        uolen = len(uo)
+        if uolen > 1:
+            candidates = short_idents
+            minName = ''
+            if uolen == 2:
+                candidates = short_idents[:53]
+            for i in candidates:
+                if i not in userobjects:
+                    minName = i
+                    userobjects[minName] = "TRN"
+                    break
+            if not minName:
+                print(f"{'Function' if otype == 'func' else 'Variable'} name {uo} could be shortened but no available "
+                      f"names found (would save {uses} bytes)")
+                continue
+                # we assume that nobody is insane enough to exhaust all *2,756* 2 character names,
+                # instead that uo is len 2 and all the 1 character names are in use (because of that we dont multiply
+                # uses by anything
+            if not auto_replace:
+                print(f"{'Function' if otype == 'func' else 'Variable'} name {uo} could be shortened ({uo}->{minName}, "
+                      f"would save {uses*(uolen - len(minName))} bytes")
+                continue
+            else:
+                print(f"Renaming {'Function' if otype == 'func' else 'Variable'} {uo} to {minName} "
+                      f"(saving {uses*(uolen - len(minName))} bytes)")
+                # rather than just blindly str.replace()ing we're going to actually use the character indices that we stored
+                diff = uolen - len(minName)
+                prev = 0
+                for bound in usages[uo]:
+                    tmpcode += mcode[prev:bound] + minName + ' '*diff
+                    prev = bound + diff + len(minName)
+                mcode = tmpcode + mcode[bound+diff+len(minName):]  # it actually cant be referenced before assignment but ok
+    for func in usages:
+        tmpcode = ""
+        candidates = short_idents
+        minName = ''
+        savings = 0
+        uses = len(usages[func])
+        if func in userobjects or uses < 2:  # we only want stdlib functions used more than once
+            continue
+        elif func == "if":
+            candidates = short_idents[:53]
+            savings = uses * 2 - 5  # the 5 is how many characters an alias declaration would use (a=if<space>)
+        for i in candidates:
+            if i not in userobjects:
+                minName = i
+                userobjects[minName] = "TRP"
+                break
+        # once again we assume it's only `if` that could trigger this message
+        if not minName and (uses - 4) > 0:
+            print(f"Standard library function {func} could be aliased but no available names found "
+                  f"(would save {uses-4} bytes)")
+        else:
+            if not savings:
+                savings = uses*len(func) - (len(func)+len(minName)+2)
+            if savings <= 0 or not auto_replace:
+                print(f"Not aliasing standard library function {func} (would save {savings} bytes)")
+            else:
+                print(f"Aliasing standard library function {func} to {minName} (saving {savings} bytes)")
+                diff = len(func) - len(minName)
+                prev = 0
+                for bound in usages[func]:
+                    tmpcode += mcode[prev:bound] + minName + ' ' * diff
+                    prev = bound + diff + len(minName)
+                mcode = tmpcode + mcode[bound + diff + len(minName):]
+                aliases.append(f"{minName}={func} ")
+
     str_reuse = {}
+    for string in script.strings:
+        if string[2] in str_reuse:
+            str_reuse[string[2]].append(string[0])
+        else:
+            str_reuse[string[2]] = [string[0]]
+    for string in str_reuse:
+        tmpcode = ""
+        candidates = short_idents
+        uses = len(str_reuse[string])
+        minName = ""
+        if uses > 1 and len(string) > 1:
+            if len(string) == 2:
+                candidates = short_idents[:53]
+            for i in candidates:
+                if i not in userobjects:
+                    minName = i
+                    userobjects[minName] = "TIV"
+                    break
+            # the quotation marks are included in string
+            savings = uses * len(string) - (len(string) + len(minName) + 2)
+            if savings <= 0 or not auto_replace:
+                print(f"Not introducing variable for string {string} reused {uses} times (would save {savings} bytes)")
+            else:
+                # "duplicated code fragment" do i look like i give a shit
+                print(f"Introducing variable {minName} with value {string} (saving {savings} bytes)")
+                diff = len(string) - len(minName)
+                prev = 0
+                for bound in str_reuse[string]:
+                    bound -= script.comments[-1][1] if script.comments else 0
+                    tmpcode += mcode[prev:bound] + minName + ' ' * diff
+                    prev = bound + diff + len(minName)
+                mcode = tmpcode + mcode[bound + diff + len(minName):]
+                aliases.append(f"{minName}={string} ")
+
+    print("Reintroducing REQUIREs")
+    mcode = "".join([x[2] for x in script.comments]) + "".join(aliases) + mcode
+    print("Stripping whitespace")
+    return whitespacent(mcode)
+
+
+def whitespacent(script: str):
+    # also happens to remove unneeded comments and push REQUIREs to the top of the file
     requires = ""
     mcode = ""
-    stl_counts = {}.fromkeys(sub_funcs, 0)
-    # while part < len(strings):
     for line in script.split(sep='\n'):
-        # maybe in future it'll shrink user defined names
+        start = hascomment(line)
-        # dont hold out hope for that because `a.files.foreach("b") {println(b)}` is valid syntax
-        # and i dont have the skill or patience to deal with that
-
-        # # in theory all the even numbered indexes should be outside quotes, so we ignore any parts with an odd index
-        # if part % 2 == 1:
-        #     if strings[part] not in str_reuse:
-        #         str_reuse[strings[part]] = 0
-        #     else:
-        #         str_reuse[strings[part]] += 1
-        #     mcode += f'"{strings[part]}"'
-        start = commentstartswhere(line)
         if start is None:
             start = -1
 
@@ -74,27 +315,20 @@ def minify(script: str):
         line = line.split(sep='"')
 
         if len(line) % 2 == 0:
-            print("You appear to have string literals spanning multiple lines. Please seek professional help")
+            raise Exception("Unmatched quote or hard newline in string")
-            raise Exception("Too much hatred")
         part = 0
         while part < len(line):
             # all the odd numbered indexes should be inside quotes
             if part % 2 == 0:
                 if not line[part]:
                     break
-                for s in sub_funcs:
-                    stl_counts[s] += len(re.findall("(?<!\\.)%s\\(" % s, line[part]))
                 mcode += line[part].replace('\t', '') + ' '
             else:
-                if line[part] not in str_reuse:
-                    str_reuse[line[part]] = 0
-                else:
-                    str_reuse[line[part]] += 1
                 mcode += f'"{line[part]}"'
 
             part += 1
 
-
+    mcode = " ".join(mcode.split())  # turn lots of whitespace into one whitespace with one easy trick!
     # tsv3 is still an absolute nightmare
     # so spaces have a couple edge cases
     # 1. the - operator which requires space between the right operand
@@ -107,12 +341,10 @@ def minify(script: str):
     newline = list(mcode)
     while index < (len(mcode) - 3):
         sec = mcode[index:index + 3]
-        if not inquote and sec[1] == '"':
+        if sec[1] == '"':
-            inquote = True
+            inquote = not inquote
-        elif inquote and sec[1] == '"':
-            inquote = False
         if (sec[1] == ' ') and not inquote:
-            if wantsumspace(sec[0]) and wantsumspace(sec[2]):
+            if (isidentifier(sec[0]) or sec[0].isnumeric()) and (isidentifier(sec[2]) or sec[2].isnumeric()):
                 pass
             elif sec[0] == '-' and sec[2].isnumeric():
                 pass
@@ -121,62 +353,29 @@ def minify(script: str):
         index += 1
     mmcode += ''.join(newline).strip()
 
-    for func in sub_funcs:
+    return requires + mmcode.strip().replace('\n', ' ')
-        # space saved here is given by n * (len(func) - len(min_func)) - (len(min_func)+1 + len(func))
-        # as such with one usage space is always lost (len(func)-2 is never > len(func)+3) so dont even try
-        if stl_counts[func] >= 2:
-            savings = stl_counts[func] * (len(func) - 2) - (len(func) + 3)
-            print(f"Replacing all {stl_counts[func]} usages of {func} would save {savings}byte{'s' if savings != 1 else ''}")
-            if (savings < 0) or not replace_functions:
-                print("Savings negative or automatic replacement disabled, continuing")
-                continue
-            func_min = sub_funcs[func]  # now here we have to assume nobody is using any of our substitute vars
-            # should be a pretty safe assumption but knowing for sure would require about the same amount of effort
-            # as it would to replace all user defined variables
-            ucode = ""  # this is rather hacky
-            sections = [0]
-            for m in re.finditer(r"(?<!\.)%s\(" % func, mmcode):
-                sections.append(m.span()[0])
-                sections.append(m.span()[1])
-            sections.append(len(mmcode))
-            i = 2       # change rather to very
-            while i < len(sections):
-                ucode += mmcode[sections[i-2]:sections[i-1]] + func_min + '('
-                i += 2
-            ucode += mmcode[sections[i-2]:]
-
-            ucode = f"{func_min}={func}\n" + ucode
-            mmcode = ucode
-            # a space isn't any shorter than \n so why not use \n
-
-    for string, count in str_reuse.items():
-        if count >= 2:
-            # we can't auto replace strings without a full parser
-            # unlike with the stdlib functions we cant make a lookup table ahead of time
-            # and generating shorter names on the fly sounds like an absolute nightmare no thanks
-            print(f'Warning: string "{string}" of len {len(string)} reused {count} times')
-
-    return requires + mmcode.strip()
 
 
 if __name__ == '__main__':
-    parser = argparse.ArgumentParser(description="Minify tsv3 scripts, useful for embedding",
+    argparser = argparse.ArgumentParser(description="Minify tsv3 scripts, useful for embedding",
                                         formatter_class=argparse.RawTextHelpFormatter)
-    parser.add_argument("source", type=str, nargs='+', help="source files to minify")
+    argparser.add_argument("source", type=str, nargs='+', help="source files to minify")
-    parser.add_argument("-d", type=str, nargs='?', help="destination folder for minified scripts"
+    argparser.add_argument("-d", type=str, nargs='?', help="destination folder for minified scripts"
                                                            "\ndefault: ./", default='./')
-    parser.add_argument("--replace-functions", action="store_true", default=False,
+    argparser.add_argument("--auto-replace", action="store_true", default=False,
-                        help="automatically replace reused functions instead of just warning\ndefault: false")
+                           help="automatically replace reused functions and variables instead of just warning\n"
+                           "and attempt to generate shorter names for reused variables \ndefault: false")
 
-    args = parser.parse_args()
+    args = argparser.parse_args()
     files = args.source
     dest = args.d[:-1] if args.d[-1] == '/' else args.d
-    replace_functions = args.replace_functions if args.replace_functions is not None else False
+    auto_replace = args.auto_replace if args.auto_replace is not None else False
 
     for file in files:
-        print(f"Minifying {file}")
+        print(f"\nMinifying {file}")
         with open(file, 'r') as f:
-            r = minify(f.read())
+            print("Stripping comments")
+            r = parser(whitespacent(f.read()))
         file = file.split(sep='.')[0].split(sep='/')[-1]
         if dest != '.':
             f = open(f"{dest}/{file}.te", 'w')