LSL-PyOptimizer/main.py
Sei Lisa 3f6f8ed8ad Internal code reorganization.
- Separate library loading code into a new module. parser.__init__() no longer loads the library; it accepts (but does not depend on) a library as a parameter.
- Add an optional library argument to parse(). It's no longer mandatory to create a new parser for switching to a different builtins or seftable file.
- Move warning() and types from lslparse to lslcommon.
- Add .copy() to uses of base_keywords, to not rely on it being a frozen set.
- Adjust the test suite.
2017-10-20 18:19:48 +02:00

694 lines
28 KiB
Python
Executable file

#!/usr/bin/env python2
# (C) Copyright 2015-2017 Sei Lisa. All rights reserved.
#
# This file is part of LSL PyOptimizer.
#
# LSL PyOptimizer is free software: you can redistribute it and/or
# modify it under the terms of the GNU General Public License as
# published by the Free Software Foundation, either version 3 of the
# License, or (at your option) any later version.
#
# LSL PyOptimizer is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with LSL PyOptimizer. If not, see <http://www.gnu.org/licenses/>.
# This is the main executable program that imports the libraries.
from lslopt.lslparse import parser,EParse
from lslopt.lsloutput import outscript
from lslopt.lsloptimizer import optimizer
import sys, os, getopt, re
import lslopt.lslcommon
import lslopt.lslloadlib
VERSION = '0.2.1beta'
def ReportError(script, e):
linestart = script.rfind(b'\n', 0, e.errorpos) + 1
lineend = script.find(b'\n', e.errorpos)
if lineend == -1: lineend = len(script) # may hit EOF
# When the encoding of stderr is unknown (e.g. when redirected to a file),
# output will be encoded in UTF-8; otherwise the terminal's encoding will
# be used.
enc = sys.stderr.encoding if sys.stderr.encoding is not None else 'utf8'
# Synchronize the UTF-8 encoded line with the output line in the
# terminal's encoding. We need to compensate for the fact that the
# reported column applies to the UTF-8 version of the script.
# 1. Trim the UTF-8 line.
err_frag = script[linestart:e.errorpos]
# 2. Convert to Unicode; encode in the target encoding with replacing.
err_frag = err_frag.decode('utf8').encode(enc, 'backslashreplace')
# 3. Collect our prize: the length of that in characters.
cno = len(err_frag.decode(enc))
# Write the whole line in the target encoding.
err_line = script[linestart:lineend] + b'\n'
sys.stderr.write(err_line.decode('utf8').encode(enc, 'backslashreplace'))
sys.stderr.write(u" " * cno + u"^\n")
sys.stderr.write(e.args[0] + u"\n")
class UniConvScript(object):
"""Converts the script to Unicode, setting the properties required by
EParse to report a meaningful error position.
"""
def __init__(self, script, options = (), filename = '<stdin>'):
self.linedir = []
self.filename = filename
# We don't interpret #line here. In case of an encode error,
# we're in the dark about which file it comes from. User needs
# --preshow to view the #line directives and find the correspondence
# themselves.
#self.processpre = 'processpre' in options
self.processpre = False
self.script = script
def to_unicode(self):
if type(self.script) is not unicode:
try:
self.script = self.script.decode('utf8')
except UnicodeDecodeError as e:
self.errorpos = e.start
raise EParse(self, u"Invalid UTF-8 in script")
return self.script
def PreparePreproc(script):
s = ''
nlines = 0
col = 0
# Trigraphs make our life really difficult.
# We join lines that have \<return> or ??/<return> inside strings,
# and we also replace regular <return> inside strings with \n, counting how
# many lines we join, to add them back at the end of the string in order to
# keep the line count exact prior to preprocessing. We also preserve the
# original column of the text after the string, by adding as many spaces as
# necessary.
# We could let the preprocessor do the line joining on backslash-newline,
# but by eliminating all newlines, we have control over the output column
# of the text that follows the string and can report an accurate column
# and line position in case of error.
# The REs skip as much as possible in one go every time, only stopping to
# analyze critical tokens.
# We don't follow the C convention that backslash-return is analyzed first.
# In c, the string "a\\<return>nb" is the same as "a\nb" which prints as
# a<return>b. But in LSL, forgetting about the preprocessor, the string
# "a\\<return>nb" is valid and stands for a\<return>nb. The principle of
# least surprise seems to suggest to accept valid LSL strings as LSL
# instead of reproducing that C quirk. This also matches what FS is doing
# currently, so it's good for compatibility.
tok = re.compile(
r'(?:'
r'/(?:\?\?/\n|\\\n)*\*.*?\*(?:\?\?/\n|\\\n)*/'
r'|/(?:\?\?/\n|\\\n)*/(?:\?\?/\n|\\\n|[^\n])*\n'
r'|[^"]'
r')+'
r'|"'
, re.S)
# RE used inside strings.
tok2 = re.compile(
r'(?:'
r"\?\?[='()!<>-]" # valid trigraph except ??/ (backslash)
r"|(?:\?\?/|\\)(?:\?\?[/='()!<>-]|[^\n])"
# backslash trigraph or actual backslash,
# followed by any trigraph or non-newline
r'|(?!\?\?/\n|\\\n|"|\n).'
# any character that doesn't start a trigraph/
# backslash escape followed by a newline
# or is a newline or double quote, as we're
# interested in all those individually.
r')+' # as many of those as possible
r'|\?\?/\n|\\\n|\n|"' # or any of those individually
)
pos = 0
match = tok.search(script, pos)
while match:
matched = match.group(0)
pos += len(matched)
if matched == '"':
s += matched
nlines = col = 0
match2 = tok2.search(script, pos)
while match2:
matched2 = match2.group(0)
pos += len(matched2)
if matched2 == '\\\n' or matched2 == '??/\n':
nlines += 1
col = 0
match2 = tok2.search(script, pos)
continue
if matched2 == '"':
if nlines:
if script[pos:pos+1] == '\n':
col = -1 # don't add spaces if not necessary
# col misses the quote added here, so add 1
s += '"' + '\n'*nlines + ' '*(col+1)
else:
s += '"'
break
if matched2 == '\n':
nlines += 1
col = 0
s += '\\n'
else:
col += len(matched2)
s += matched2
match2 = tok2.search(script, pos)
else:
s += matched
match = tok.search(script, pos)
return s
def ScriptHeader(script, avname):
if avname:
avname = ' - ' + avname
return ('//start_unprocessed_text\n/*'
# + re.sub(r'([*/])(?=[*|/])', r'\1|', script) # FS's algorithm
# HACK: This won't break strings containing ** or /* or // like URLs,
# while still being compatible with FS.
+ re.sub(r'([*/]\||\*(?=/))', r'\1|', script)
+ '*/\n//end_unprocessed_text\n//nfo_preprocessor_version 0\n'
'//program_version LSL PyOptimizer v' + VERSION + avname
+ '\n//mono\n\n')
def Usage(progname, about = None):
if about is None:
sys.stderr.write(
ur"""LSL optimizer v{version}
(C) Copyright 2015 Sei Lisa. All rights reserved.
This program comes with ABSOLUTELY NO WARRANTY.
This is free software, and you are welcome to redistribute it
under certain conditions; see the file COPYING for details.
This program is licensed under the GNU General Public License
version 3.
Usage: {progname}
[-O|--optimizer-options=[+|-]option[,[+|-]option[,...]]]
optimizer options (use '-O help' for help)
[-h|--help] print this help
[--version] print this program's version
[-o|--output=<filename>] output to file rather than stdout
[-b|--builtins=<filename>] use a builtins file other than builtins.txt
[-S|--seftable=<filename>] use a SEF table file other than seftable.txt
[-H|--header] add the script as a comment in Firestorm format
[-T|--timestamp] add a timestamp as a comment at the beginning
[-y|--python-exceptions] when an exception is raised, show a stack trace
[-p|--preproc=mode] run external preprocessor (see below for modes)
(resets the preprocessor command line so far)
[-P|--prearg=<arg>] add parameter to preprocessor's command line
[--precmd=<cmd>] preprocessor command ('cpp' by default)
[--prenodef] no LSL specific defines (__AGENTKEY__ etc.)
[--preshow] show preprocessor output, and stop
[--avid=<UUID>] * specify UUID of avatar saving the script
[--avname=<name>] * specify name of avatar saving the script
[--assetid=<UUID>] * specify the asset UUID of the script
[--shortname=<name>] * specify the script's short file name
filename input file
Options marked with * are used to define the preprocessor macros __AGENTID__,
__AGENTKEY__, __AGENTIDRAW__, __AGENTNAME__, __ASSETID__ and __SHORTFILE__,
and have no effect if --prenodef is specified.
Using --prenodef before -p causes no macros whatsoever to be defined. If used
after -p, the type casting macros string(...), key(...), etc. will be defined.
If filename is a dash (-) then standard input is used.
Use: {progname} -O help for help on the optimizer control options.
Preprocessor modes:
ext Invoke preprocessor
mcpp Invoke mcpp as preprocessor, setting default parameters pertinent
to it. Implies --precmd=mcpp
gcpp Invoke GNU cpp as preprocessor, setting default parameters
pertinent to it. Implies --precmd=cpp
none No preprocessing (default)
Normally, running the preprocessor needs the option 'processpre' active, to
make the output readable by the optimizer. This option is active by default.
""".format(progname=progname, version=VERSION))
return
if about == 'optimizer-options':
sys.stderr.write(
ur"""
Optimizer control options.
+ means active by default, - means inactive by default.
Case insensitive.
Syntax extensions options:
ExtendedGlobalExpr + Enables arbitrary expressions in globals (as opposed to
dull simple expressions allowed by regular LSL). Needs
constant folding active for the result to be compilable.
BreakCont - Allow break/continue statements for loops. Note that
when active, 'break' and 'continue' become reserved
words, but when inactive they can be used as variables.
ExtendedTypeCast + Allows extended typecast syntax e.g. (string)(integer)a
is valid with this option.
ExtendedAssignment + Enables &=, |=, ^=, <<=, >>= assignment operators.
AllowKeyConcat + Allow string + key and key + string (both return string)
AllowMultiStrings + Allow C-like string juxtaposition, e.g. "ab" "cd" means
"abcd", no concatenation involved. Very useful when used
with a preprocessor. Similar to addstrings, but this one
is not an optimization, it introduces new syntax.
DupLabels - Normally, a duplicate label within a function is allowed
by the syntax by using {{}} blocks; however, the server
will just refuse to save the script (under Mono) or do
something completely unexpected (under LSO: only the
last jump will execute, and it will go to the last label
with that name). This flag works around that limitation
by replacing the names of the labels in the output with
unique ones.
Deprecated / compatibility syntax extensions options:
LazyLists - Support syntax like mylist[index] = 5; rather than using
llListReplaceList. Only assignment supported. The list
is extended when the argument is greater than the list
length, by inserting integer zeros. This is implemented
for compatibility with Firestorm, but its use is not
recommended, as it adds a new function, wasting memory
against the very spirit of this program.
EnableSwitch - Support C-like switch() syntax, with some limitations.
Like lazylists, it's implemented for compatibility with
Firestorm, but not recommended. Note that the operand to
switch() may be evaluated more than once.
ErrMissingDefault + Throw an error in case the 'default:' label of a switch
statement is missing. Only meaningful with EnableSwitch.
FuncOverride - Allow duplicate function definitions to override the
previous definition. For compatibility with Firestorm's
optimizer.
Optimization options
Optimize + Runs the optimizer.
OptSigns + Optimize signs in float and integer constants.
OptFloats + Optimize floats that represent an integral value.
ConstFold + Fold constant expressions to their values, and simplify
some expressions and statements.
DCR + Dead code removal. This option removes several instances
of code that will never execute, and performs other
optimizations like removal of unused variables,
functions or expressions.
ShrinkNames - Reduces script memory by shrinking identifiers. In the
process, it turns the script into unreadable gibberish,
hard to debug, but this gets big savings for complex
scripts.
AddStrings - Concatenate strings together when possible. Note that
such an optimization can be counter-productive in some
cases, that's why it's disabled by default. For example:
string a="a"+"longstring"; string b="b"+"longstring";
would keep a single copy of "longstring", while if the
strings are added, both "alongstring" and "blongstring"
take memory.
ListLength + Optimize llGetListLength(arg) to arg!=[]. Needs constant
folding active to work.
ListAdd + Convert [a,b,c...] to (list)a + b + c... if possible.
Miscellaneous options
FoldTabs - Tabs can't be copy-pasted, so expressions that produce
tabs, e.g. llUnescapeURL("%09"), aren't optimized by
default. This option overrides that check, enabling
expansion of functions that produce strings with tabs.
The resulting source isn't guaranteed to be
copy-paste-able to the viewer.
WarnTabs + Warn when a function can't be optimized because it
generates a string or list with a tab, or when a string
contains a tab.
ProcessPre + Process some preprocessor directives in the source. This
enables usage of #pragma/#line preprocessor directives,
and is probably necessary if the script is itself the
output of a preprocessor. Note that this option does not
make the optimizer process macros.
ExplicitCast - Add explicit casts where they are implicit. This option
is useless with 'optimize' and 'optsigns', and is of
basically no use in general, other than to see where
automatic casts happen.
Clear - Set all options to inactive, Normally used as the first
option, to start afresh. Note that this sets to inactive
even the options that are active by default.
For example:
{progname} -O -DCR,+BreakCont scriptname.lsl
would turn off dead code removal (which is active by default) and turn on the
break/continue syntax extension (which is inactive by default).
""".format(progname=progname))
return
validoptions = frozenset(('extendedglobalexpr','breakcont','extendedtypecast',
'extendedassignment','allowkeyconcat','allowmultistrings','duplabels',
'lazylists','enableswitch','errmissingdefault','funcoverride','optimize',
'optsigns','optfloats','constfold','dcr','shrinknames','addstrings',
'foldtabs','warntabs','processpre','explicitcast','listlength','listadd',
'help',
# undocumented
'lso','expr','rsrclimit',
# 'clear' is handled as a special case
))
def main(argv):
"""Main executable."""
# If it's good to append the basename to it, it's good to append the
# auxiliary files' names to it, which should be located where this file is.
lslopt.lslcommon.DataPath = __file__[:-len(os.path.basename(__file__))]
# Default options
options = set(('extendedglobalexpr','extendedtypecast','extendedassignment',
'allowkeyconcat','allowmultistrings','processpre','warntabs','optimize',
'optsigns','optfloats','constfold','dcr','errmissingdefault',
'listlength','listadd',
))
assert not (options - validoptions), (u"Default options not present in"
u" validoptions: '%s'"
% (b"', '".join(options - validoptions)).decode('utf8'))
try:
opts, args = getopt.gnu_getopt(argv[1:], 'hO:o:p:P:HTyb:S:',
('optimizer-options=', 'help', 'version', 'output=', 'header',
'timestamp','python-exceptions',
'preproc=', 'precmd=', 'prearg=', 'prenodef', 'preshow',
'avid=', 'avname=', 'assetid=', 'shortname=', 'builtins='
'seftable='))
except getopt.GetoptError as e:
Usage(argv[0])
sys.stderr.write(u"\nError: " + str(e).decode('utf8') + u"\n")
return 1
outfile = '-'
avid = '00000000-0000-0000-0000-000000000000'
avname = ''
shortname = ''
assetid = '00000000-0000-0000-0000-000000000000'
preproc_cmdline = ['cpp']
preproc = 'none'
predefines = True
script_header = ''
script_timestamp = ''
mcpp_mode = False
preshow = False
raise_exception = False
builtins = None
seftable = None
for opt, arg in opts:
if type(opt) is unicode:
opt = opt.encode('utf8')
if type(arg) is unicode:
arg = arg.encode('utf8')
if opt in ('-O', '--optimizer-options'):
optchanges = arg.lower().split(',')
for chg in optchanges:
if not chg:
continue
if chg in ('clear', '+clear'):
options = set()
continue
if chg == '-clear':
# ignore
continue
chgfix = chg
if chgfix[0] not in ('+', '-'):
chgfix = '+' + chgfix
if chgfix[1:] not in validoptions:
Usage(argv[0], 'optimizer-options')
sys.stderr.write(u"\nError: Unrecognized"
u" optimizer option: %s\n" % chg.decode('utf8'))
return 1
if chgfix[0] == '-':
options.discard(chgfix[1:])
else:
options.add(chgfix[1:])
del chgfix
elif opt in ('-h', '--help'):
Usage(argv[0])
return 0
elif opt == '--version':
sys.stdout.write('LSL PyOptimizer version %s\n' % VERSION)
return 0
elif opt in ('-o', '--output'):
outfile = arg
elif opt in ('-b', '--builtins'):
builtins = arg
elif opt in ('-S', '--seftable'):
seftable = arg
elif opt in ('-y', '--python-exceptions'):
raise_exception = True
elif opt in ('-p', '--preproc'):
preproc = arg.lower()
supported = ('ext', 'mcpp', 'gcpp', 'none')
if preproc not in supported:
Usage(argv[0])
sys.stderr.write(u"\nUnknown --preproc option: '%s'."
u" Only '%s' supported.\n"
% (preproc, u"', '".join(supported)))
return 1
mcpp_mode = False
del preproc_cmdline[1:]
if preproc == 'gcpp':
preproc_cmdline = [
'cpp', '-undef', '-x', 'c', '-std=c99', '-nostdinc',
'-trigraphs', '-dN', '-fno-extended-identifiers',
]
elif preproc == 'mcpp':
mcpp_mode = True
preproc_cmdline = [
'mcpp', '-e', 'UTF-8', '-I-', '-N', '-3', '-j',
'-V199901L',
]
if predefines:
preproc_cmdline += [
'-Dinteger(...)=((integer)(__VA_ARGS__))',
'-Dfloat(...)=((float)(__VA_ARGS__))',
'-Dstring(...)=((string)(__VA_ARGS__))',
'-Dkey(...)=((key)(__VA_ARGS__))',
'-Drotation(...)=((rotation)(__VA_ARGS__))',
'-Dquaternion(...)=((quaternion)(__VA_ARGS__))',
'-Dvector(...)=((vector)(__VA_ARGS__))',
'-Dlist(...)=((list)(__VA_ARGS__))',
]
elif opt == '--precmd':
preproc_cmdline[0] = arg
elif opt in ('-P', '--prearg'):
preproc_cmdline.append(arg)
elif opt == '--prenodef':
predefines = False
elif opt == '--preshow':
preshow = True
elif opt in ('-H', '--header'):
script_header = True
elif opt in ('-T', '--timestamp'):
script_timestamp = True
elif opt == '--avid':
avid = arg
elif opt == '--avname':
avname = arg
elif opt == '--assetid':
assetid = arg
elif opt == '--shortname':
shortname = arg
del opts
try:
if 'rsrclimit' in options:
import resource
resource.setrlimit(resource.RLIMIT_CPU, (5, 5))
resource.setrlimit(resource.RLIMIT_STACK, (393216, 393216))
resource.setrlimit(resource.RLIMIT_DATA, (4096, 4096))
resource.setrlimit(resource.RLIMIT_AS, (20001000, 20001000))
if 'lso' in options:
lslopt.lslcommon.LSO = True
options.remove('lso')
if 'expr' in options:
lslopt.lslcommon.IsCalc = True
options.remove('expr')
if 'help' in options:
Usage(argv[0], 'optimizer-options')
return 0
fname = args[0] if args else None
if fname is None:
Usage(argv[0])
sys.stderr.write(u"\nError: Input file not specified. Use -"
u" if you want to use stdin.\n")
return 1
del args
script = ''
if fname == '-':
script = sys.stdin.read()
else:
try:
f = open(fname, 'r')
except IOError as e:
if e.errno == 2:
sys.stderr.write('Error: File not found: %s\n' % fname)
return 2
raise
try:
script = f.read()
finally:
f.close()
del f
if script_header:
script_header = ScriptHeader(script, avname)
if script_timestamp:
import time
tmp = time.time()
script_timestamp = time.strftime(
'// Generated on %Y-%m-%dT%H:%M:%S.{0:06d}Z\n'
.format(int(tmp % 1 * 1000000)), time.gmtime(tmp))
del tmp
if shortname == '':
shortname = os.path.basename(fname)
if predefines:
preproc_cmdline.append('-D__AGENTKEY__="' + avid + '"')
preproc_cmdline.append('-D__AGENTID__="' + avid + '"')
preproc_cmdline.append('-D__AGENTIDRAW__=' + avid)
preproc_cmdline.append('-D__AGENTNAME__="' + avname + '"')
preproc_cmdline.append('-D__ASSETID__=' + assetid)
preproc_cmdline.append('-D__SHORTFILE__="' + shortname + '"')
preproc_cmdline.append('-D__OPTIMIZER__=LSL PyOptimizer')
preproc_cmdline.append('-D__OPTIMIZER_VERSION__=' + VERSION)
if type(script) is unicode:
script = script.encode('utf8')
else:
try:
# Try converting the script to Unicode, to report any encoding
# errors with accurate line information. At this point we don't
# need the result.
UniConvScript(script, options,
fname if fname != '-' else '<stdin>').to_unicode()
except EParse as e:
# We don't call ReportError to prevent problems due to
# displaying invalid UTF-8
sys.stderr.write(e.args[0] + u"\n")
return 1
if preproc != 'none':
# At this point, for the external preprocessor to work we need the
# script as a byte array, not as unicode, but it should be UTF-8.
script = PreparePreproc(script)
if mcpp_mode:
# As a special treatment for mcpp, we force it to output its
# macros so we can read if USE_xxx are defined. With GCC that
# is achieved with -dN, but mcpp has no command line option.
script += '\n#pragma MCPP put_defines\n'
# Invoke the external preprocessor
import subprocess
p = subprocess.Popen(preproc_cmdline, stdin=subprocess.PIPE,
stdout=subprocess.PIPE)
script = p.communicate(input=script)[0]
status = p.wait()
if status:
return status
del p, status
# This method is very imperfect, in several senses. However, since
# it's applied to the output of the preprocessor, all of the
# concerns should be addressed:
# - \s includes \n, but \n should not be allowed.
# - Comments preceding the directive should not cause problems.
# e.g.: /* test */ #directive
# - #directive within a comment or string should be ignored.
for x in re.findall(r'(?:(?<=\n)|^)\s*#\s*define\s+('
r'USE_SWITCHES'
r'|USE_LAZY_LISTS'
r')(?:$|[^A-Za-z0-9_])', script, re.S):
if x == 'USE_SWITCHES':
options.add('enableswitch')
elif x == 'USE_LAZY_LISTS':
options.add('lazylists')
if not preshow:
lib = lslopt.lslloadlib.LoadLibrary(builtins, seftable)
p = parser(lib)
try:
ts = p.parse(script, options,
fname if fname != '-' else '<stdin>')
except EParse as e:
ReportError(script, e)
return 1
del p, script
opt = optimizer()
ts = opt.optimize(ts, options)
del opt
outs = outscript()
script = script_header + script_timestamp + outs.output(ts, options)
del outs, ts
del script_header, script_timestamp
if outfile == '-':
sys.stdout.write(script)
else:
outf = open(outfile, 'w')
try:
outf.write(script)
finally:
outf.close()
return 0
except Exception as e:
if raise_exception:
raise
sys.stderr.write(e.__class__.__name__ + ': ' + str(e) + '\n')
return 1
if __name__ == '__main__':
ret = main(sys.argv)
if ret:
sys.exit(ret)