mirror of
https://github.com/Sei-Lisa/LSL-PyOptimizer
synced 2025-07-01 23:58:20 +00:00
075d3aba0c
That was long overdue. Obviously, this is a large commit. The new nr (node record) class has built-in dump capabilities, rather than using print_node(). SEF always exists now, and is a boolean, rather than using the existence of SEF as the flag. This was changed for sanity. However, other flags like 'X' are still possibly absent, and in some cases the absence itself has meaning (in the case of 'X', its absence means that the node has not yet been analyzed). Similarly, an event is distinguished from a UDF by checking for the existence of the 'scope' attribute. This trick works because events are not in the symbol table therefore they have no scope. But this should probably be changed in future to something more rational and faster. A few minor bugfixes were applied while going through the code. - Some tabs used as Unicode were written as byte strings. Add the u'\t' prefix. - After simplifying a%1 -> a&0, fold again the node and return. It's not clear why it didn't return, and whether it depended on subsequent passes (e.g. after DCR) for possibly optimizing out the result. Now we're sure. - A few places lacked a SEF declaration. - Formatting changes to split lines that spilled the margin. - Some comment changes. - Expanded lazy_list_set definition while adapting it to object format. The plan was to re-compress it after done, but decided to leave it in expanded form. - Added a few TODOs & FIXMEs, resisting the temptation to fix them in the same commit: - TODO: ~-~-~-expr -> expr + -3. - FIXME: Now that we have CompareTrees, we can easily check if expr + -expr cancels out and remove a TODO. Low-hanging fruit. - TODO: Check what we can do when comparing non-SEF and non-CONST values in '>' (current code relies on converting '>' to '<' for applying more optimizations, but that may miss some opportunities). - FIXME: Could remove one comparison in nt == '&&' or nt == '||'. Low-hanging fruit.
507 lines
21 KiB
Python
507 lines
21 KiB
Python
# (C) Copyright 2015-2018 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/>.
|
|
|
|
# Convert an abstract syntax tree + symbol table back to a script as text.
|
|
|
|
import lslfuncs
|
|
import lslcommon
|
|
from lslcommon import Key, Vector, Quaternion, warning
|
|
from math import copysign
|
|
|
|
class outscript(object):
|
|
|
|
binary_operands = frozenset(('||','&&','^','|','&','==','!=','<','<=','>',
|
|
'>=','<<','>>','+','-','*','/','%', '=', '+=', '-=', '*=', '/=','%=',
|
|
))
|
|
extended_assignments = frozenset(('&=', '|=', '^=', '<<=', '>>='))
|
|
unary_operands = frozenset(('NEG', '!', '~'))
|
|
op_priority = {'=':0, '+=':0, '-=':0, '*=':0, '/=':0, '%=':0, '&=':0,
|
|
'|=':0, '^=':0, '<<=':0, '>>=':0,
|
|
'||':1, '&&':1, '|':2, '^':3, '&':4, '==':5, '!=':5,
|
|
'<':6, '<=':6, '>':6, '>=':6, '<<':7, '>>':7, '+':8, '-':8,# 'NEG':8,
|
|
'*':9, '/':9, '%':9}#, '!':10, '~':10, '++':10, '--':10, }
|
|
assignment_ops = ('=', '+=', '-=', '*=', '/=','%=')
|
|
|
|
def Value2LSL(self, value):
|
|
tvalue = type(value)
|
|
if tvalue in (Key, unicode):
|
|
pfx = sfx = ''
|
|
if type(value) == Key:
|
|
# Constants of type key can not be represented
|
|
#raise lslfuncs.ELSLTypeMismatch
|
|
# Actually they can be the result of folding.
|
|
# On second thought, if we report the error, the location info
|
|
# is lost. So we emit a warning instead, letting the compiler
|
|
# report the error in the generated source.
|
|
if self.globalmode and self.listmode:
|
|
warning(u"Illegal combo: Key type inside a global list")
|
|
if self.listmode or not self.globalmode:
|
|
if self.globalmode:
|
|
pfx = '(key)'
|
|
else:
|
|
pfx = '((key)'
|
|
sfx = ')'
|
|
if u'\t' in value and self.warntabs:
|
|
warning(u"A string contains a tab. Tabs are expanded to four"
|
|
" spaces by the viewer when copy-pasting the code"
|
|
" (disable this warning by disabling the 'warntabs'"
|
|
" option).")
|
|
return pfx + '"' + value.encode('utf8').replace('\\','\\\\') \
|
|
.replace('"','\\"').replace('\n','\\n') + '"' + sfx
|
|
if tvalue == int:
|
|
if value < 0 and not self.globalmode and self.optsigns:
|
|
#return '0x%X' % (value + 4294967296)
|
|
return '((integer)' + str(value) + ')'
|
|
return str(value)
|
|
if tvalue == float:
|
|
if self.optfloats and value.is_integer() and -2147483648.0 <= value < 2147483648.0:
|
|
if self.globalmode and not self.listmode:
|
|
if value == 0 and copysign(1, value) == -1:
|
|
return '-0.'
|
|
return str(int(value))
|
|
elif not self.globalmode:
|
|
# Important inside lists!!
|
|
if value == 0 and copysign(1, value) == -1:
|
|
return '(-(float)0)'
|
|
return '((float)' + str(int(value)) + ')'
|
|
s = repr(value)
|
|
if s == 'nan':
|
|
return '(1e40*0)' if copysign(1, value) < 0 else '(-1e40*0)'
|
|
if s == 'inf':
|
|
return '1e40'
|
|
if s == '-inf':
|
|
return '-1e40' if self.globalmode else '((float)-1e40)'
|
|
# Try to remove as many decimals as possible but keeping the F32 value intact
|
|
exp = s.find('e')
|
|
if ~exp:
|
|
s, exp = s[:exp], s[exp:]
|
|
if exp[1] == '+':
|
|
exp = exp[:1] + exp[2:]
|
|
if '.' not in s:
|
|
# I couldn't produce one but it's assumed that if it happens,
|
|
# this code deals with it correctly
|
|
s += '.' # pragma: no cover
|
|
else:
|
|
if '.' not in s:
|
|
# This should never happen (Python should always return a point or exponent)
|
|
return s + '.' # pragma: no cover
|
|
exp = ''
|
|
|
|
# Shorten the float as much as possible.
|
|
while s[-1] != '.' and lslfuncs.F32(float(s[:-1]+exp)) == value:
|
|
s = s[:-1]
|
|
if s[-1] != '.':
|
|
news = s[:-1]
|
|
neg = ''
|
|
if s[0] == '-':
|
|
news = news[1:]
|
|
neg = '-'
|
|
# Try harder
|
|
point = news.index('.') + 1 - len(news)
|
|
if point:
|
|
news = str(int(news[:point-1] + news[point:]) + 1).zfill(len(news)-1) # Increment
|
|
else:
|
|
news = str(int(news[:-1])+1).zfill(len(news)-1)
|
|
news = news[:point + len(news)] + '.' + news[point + len(news):] # Reinsert point
|
|
# Repeat the operation with the incremented number
|
|
while news[-1] != '.' and lslfuncs.F32(float(neg+news[:-1]+exp)) == value:
|
|
news = news[:-1]
|
|
if len(neg+news) < len(s) and lslfuncs.F32(float(neg+news+exp)) == value:
|
|
# Success! But we try even harder. We may have converted
|
|
# 9.9999e3 into 10.e3; that needs to be turned into 1.e4.
|
|
if exp != '':
|
|
if news[2:3] == '.': # we converted 9.9... into 10.
|
|
newexp = 'e' + str(int(exp[1:])+1) # increase exponent
|
|
news2 = news[0] + '.' + news[1] + news[3:] # move dot to the left
|
|
while news2[-1] == '0': # remove trailing zeros
|
|
news2 = news2[:-1]
|
|
if len(neg+news2) < len(s) and lslfuncs.F32(float(neg+news2+newexp)) == value:
|
|
news = news2
|
|
exp = newexp
|
|
s = neg+news
|
|
if exp and s[-1] == '.':
|
|
s = s[:-1] # transfrom e.g. 1.e-30 into 1e-30
|
|
if value >= 0 or self.globalmode or not self.optsigns:
|
|
return s + exp
|
|
return '((float)' + s + exp + ')'
|
|
if tvalue == Vector:
|
|
return '<' + self.Value2LSL(value[0]) + ', ' + self.Value2LSL(value[1]) \
|
|
+ ', ' + self.Value2LSL(value[2]) + '>'
|
|
if tvalue == Quaternion:
|
|
return '<' + self.Value2LSL(value[0]) + ', ' + self.Value2LSL(value[1]) \
|
|
+ ', ' + self.Value2LSL(value[2]) + ', ' + self.Value2LSL(value[3]) + '>'
|
|
if tvalue == list:
|
|
if value == []:
|
|
return '[]'
|
|
if len(value) < 5:
|
|
save_listmode = self.listmode
|
|
self.listmode = True
|
|
ret = '[' + self.Value2LSL(value[0])
|
|
for elem in value[1:]:
|
|
ret += ', ' + self.Value2LSL(elem)
|
|
ret += ']'
|
|
self.listmode = save_listmode
|
|
return ret
|
|
ret = '' if lslcommon.IsCalc else '\n'
|
|
first = True
|
|
self.indentlevel += 0 if lslcommon.IsCalc else 1
|
|
for entry in value:
|
|
ret += self.dent() + ('[ ' if first else ', ')
|
|
save_listmode = self.listmode
|
|
self.listmode = True
|
|
ret += self.Value2LSL(entry) + '\n'
|
|
self.listmode = save_listmode
|
|
first = False
|
|
ret += self.dent()
|
|
self.indentlevel -= 0 if lslcommon.IsCalc else 1
|
|
return ret + ']'
|
|
|
|
assert False, u'Value of unknown type in Value2LSL: ' + repr(value)
|
|
|
|
def dent(self):
|
|
return self.indent * self.indentlevel
|
|
|
|
def FindName(self, node, scope = None):
|
|
if scope is None:
|
|
# node is a node
|
|
if (hasattr(node, 'scope')
|
|
and 'NewName' in self.symtab[node.scope][node.name]):
|
|
return self.symtab[node.scope][node.name]['NewName']
|
|
if node.nt == 'FNCALL' and 'NewName' in self.symtab[0][node.name]:
|
|
return self.symtab[0][node.name]['NewName']
|
|
|
|
return node.name
|
|
# node is a name
|
|
if 'NewName' in self.symtab[scope][node]:
|
|
return self.symtab[scope][node]['NewName']
|
|
return node
|
|
|
|
def OutIndented(self, node):
|
|
if node.nt != '{}':
|
|
self.indentlevel += 1
|
|
ret = self.OutCode(node)
|
|
if node.nt != '{}':
|
|
self.indentlevel -= 1
|
|
return ret
|
|
|
|
def OutExprList(self, L):
|
|
ret = ''
|
|
if L:
|
|
First = True
|
|
for item in L:
|
|
if not First:
|
|
ret += ', '
|
|
ret += self.OutExpr(item)
|
|
First = False
|
|
return ret
|
|
|
|
def OutExpr(self, expr):
|
|
# Handles expression nodes (as opposed to statement nodes)
|
|
nt = expr.nt
|
|
child = expr.ch
|
|
|
|
if nt in self.binary_operands:
|
|
lnt = child[0].nt
|
|
lparen = False
|
|
rnt = child[1].nt
|
|
rparen = False
|
|
if nt in self.assignment_ops and nt in self.op_priority:
|
|
# Assignment is right-associative, so it needs to be dealt with
|
|
# separately.
|
|
base_pri = self.op_priority[nt]
|
|
if rnt in self.op_priority:
|
|
if self.op_priority[rnt] < base_pri: # should never happen
|
|
rparen = True
|
|
elif nt in self.op_priority:
|
|
base_pri = self.op_priority[nt]
|
|
if lnt in self.op_priority:
|
|
if self.op_priority[lnt] < base_pri:
|
|
lparen = True
|
|
elif lnt == 'NEG' and base_pri > self.op_priority['-']:
|
|
lparen = True
|
|
|
|
# This situation has ugly cases due to the strange precedence
|
|
# of unary minus. Consider the following two statements:
|
|
# (~-a) * a
|
|
# a * (~-a) * a
|
|
# In one case, the (~-a) is a left child; in the other, it's
|
|
# part of a right child. In both, cases, the parentheses are
|
|
# mandatory, or they would be interpreted respectively as:
|
|
# ~-(a * a)
|
|
# a * ~-(a * a)
|
|
# Yet the tree structure makes it quite hard to detect these.
|
|
# So as a safeguard, for now we parenthesize all ~ and ! within
|
|
# binary operands, as they have a deceitful binding power when
|
|
# there's a unary minus downstream.
|
|
#
|
|
# TODO: See if the parenthesizing of ~ and ! can be improved.
|
|
elif lnt in ('~', '!'):
|
|
lparen = True
|
|
|
|
if rnt in self.op_priority:
|
|
if self.op_priority[rnt] <= base_pri:
|
|
rparen = True
|
|
# see above
|
|
elif rnt in ('~', '!'):
|
|
rparen = True
|
|
|
|
if lparen:
|
|
ret = '(' + self.OutExpr(child[0]) + ')'
|
|
else:
|
|
ret = self.OutExpr(child[0])
|
|
ret += ' ' + nt + ' '
|
|
if rparen:
|
|
ret += '(' + self.OutExpr(child[1]) + ')'
|
|
else:
|
|
ret += self.OutExpr(child[1])
|
|
return ret
|
|
|
|
if nt == 'IDENT':
|
|
return self.FindName(expr)
|
|
|
|
if nt == 'CONST':
|
|
if (self.foldconst and expr.t == 'list' and len(expr.value) == 1
|
|
and not self.globalmode):
|
|
return '(list)' + self.Value2LSL(expr.value[0])
|
|
return self.Value2LSL(expr.value)
|
|
|
|
if nt == 'CAST' or self.foldconst and nt in ('LIST', 'CONST') and len(child)==1 and not self.globalmode:
|
|
ret = '(' + expr.t + ')'
|
|
expr = child[0]
|
|
if expr.nt in ('CONST', 'IDENT', 'V++', 'V--', 'VECTOR',
|
|
'ROTATION', 'LIST', 'FIELD', 'PRINT', 'FNCALL'):
|
|
if expr.nt != 'LIST' or len(expr.ch) != 1:
|
|
return ret + self.OutExpr(expr)
|
|
return ret + '(' + self.OutExpr(expr) + ')'
|
|
|
|
if nt == 'LIST':
|
|
self.listmode = True
|
|
if len(child) < 5:
|
|
ret = '[' + self.OutExprList(child) + ']'
|
|
else:
|
|
self.indentlevel += 0 if lslcommon.IsCalc else 1
|
|
ret = '' if lslcommon.IsCalc else '\n'
|
|
first = True
|
|
for elem in child:
|
|
ret += self.dent() + ('[ ' if first else ', ')
|
|
ret += self.OutExpr(elem) + '\n'
|
|
first = False
|
|
ret += self.dent() + ']'
|
|
self.indentlevel -= 0 if lslcommon.IsCalc else 1
|
|
self.listmode = False
|
|
return ret
|
|
|
|
if nt in ('VECTOR', 'ROTATION'):
|
|
ret = ('<' + self.OutExpr(child[0]) + ','
|
|
+ self.OutExpr(child[1]) + ',')
|
|
if nt == 'ROTATION':
|
|
ret += self.OutExpr(child[2]) + ','
|
|
lnt = child[-1].nt
|
|
if lnt in self.op_priority \
|
|
and self.op_priority[lnt] <= self.op_priority['>']:
|
|
ret += '(' + self.OutExpr(child[-1]) + ')'
|
|
else:
|
|
ret += self.OutExpr(child[-1])
|
|
return ret + '>'
|
|
|
|
if nt == 'FNCALL':
|
|
return self.FindName(expr) + '(' + self.OutExprList(child) + ')'
|
|
|
|
if nt == 'PRINT':
|
|
return 'print(' + self.OutExpr(child[0]) + ')'
|
|
|
|
if nt in self.unary_operands:
|
|
ret = nt
|
|
lnt = child[0].nt
|
|
paren = False
|
|
if nt == 'NEG':
|
|
ret = '-'
|
|
if (lnt == 'CONST' and child[0].t == 'integer'
|
|
and child[0].value < 0
|
|
):
|
|
# shortcut
|
|
ret += str(child[0].value + 4294967296)
|
|
return ret
|
|
if lnt in self.op_priority:
|
|
paren = self.op_priority[lnt] <= self.op_priority['-']
|
|
elif (lnt == 'NEG' or lnt == '--V'
|
|
or lnt == 'CONST'
|
|
and child[0].t == 'float'
|
|
and child[0].value < 0
|
|
):
|
|
ret += ' ' # don't output "--" as that's a different token
|
|
else:
|
|
if lnt in self.op_priority:
|
|
paren = True
|
|
|
|
if paren:
|
|
ret += '(' + self.OutExpr(child[0]) + ')'
|
|
else:
|
|
ret += self.OutExpr(child[0])
|
|
return ret
|
|
|
|
if nt == 'FLD':
|
|
return self.OutExpr(child[0]) + '.' + expr.fld
|
|
|
|
if nt in ('V--', 'V++'):
|
|
return self.OutExpr(child[0]) + ('++' if nt == 'V++' else '--')
|
|
|
|
if nt in ('--V', '++V'):
|
|
return ('++' if nt == '++V' else '--') + self.OutExpr(child[0])
|
|
|
|
if nt in self.extended_assignments:
|
|
lvalue = self.OutExpr(child[0])
|
|
return lvalue + ' = ' + lvalue + ' ' + nt[:-1] + ' (' + self.OutExpr(child[1]) + ')'
|
|
|
|
if nt == 'EXPRLIST':
|
|
return self.OutExprList(child)
|
|
|
|
if nt == 'SUBIDX':
|
|
return '(MISSING TYPE)' + self.OutExpr(child[0]) + '[' + self.OutExprList(child[1:]) + ']'
|
|
|
|
assert False, 'Internal error: expression type "' + nt + '" not handled' # pragma: no cover
|
|
|
|
def OutCode(self, node):
|
|
nt = node.nt
|
|
child = node.ch
|
|
|
|
if nt == 'IF':
|
|
ret = self.dent()
|
|
while True:
|
|
ret += 'if (' + self.OutExpr(child[0]) + ')\n'
|
|
# Do we need to add braces around the THEN side?
|
|
needs_braces = False
|
|
if len(child) == 3:
|
|
testnode = child[1]
|
|
# Find last IF in an ELSE IF chain
|
|
while testnode.nt == 'IF' and len(testnode.ch) == 3:
|
|
testnode = testnode.ch[2]
|
|
if testnode.nt == 'IF':
|
|
# hit an IF without ELSE at the end of the chain
|
|
needs_braces = True
|
|
if needs_braces:
|
|
ret += self.dent() + '{\n'
|
|
ret += self.OutIndented(child[1])
|
|
ret += self.dent() + '}\n'
|
|
else:
|
|
ret += self.OutIndented(child[1])
|
|
if len(child) < 3:
|
|
return ret
|
|
if child[2].nt != 'IF':
|
|
ret += self.dent() + 'else\n' + self.OutIndented(child[2])
|
|
return ret
|
|
ret += self.dent() + 'else '
|
|
node = child[2]
|
|
child = node.ch
|
|
if nt == 'WHILE':
|
|
ret = self.dent() + 'while (' + self.OutExpr(child[0]) + ')\n'
|
|
ret += self.OutIndented(child[1])
|
|
return ret
|
|
if nt == 'DO':
|
|
ret = self.dent() + 'do\n'
|
|
ret += self.OutIndented(child[0])
|
|
return ret + self.dent() + 'while (' + self.OutExpr(child[1]) + ');\n'
|
|
if nt == 'FOR':
|
|
ret = self.dent() + 'for ('
|
|
ret += self.OutExpr(child[0])
|
|
ret += '; ' + self.OutExpr(child[1]) + '; '
|
|
ret += self.OutExpr(child[2])
|
|
ret += ')\n'
|
|
ret += self.OutIndented(child[3])
|
|
return ret
|
|
if nt == '@':
|
|
return self.dent() + '@' + self.FindName(node) + ';\n'
|
|
if nt == 'JUMP':
|
|
return self.dent() + 'jump ' + self.FindName(node) + ';\n'
|
|
if nt == 'STSW':
|
|
return self.dent() + 'state ' + self.FindName(node) + ';\n'
|
|
if nt == 'RETURN':
|
|
if child:
|
|
return self.dent() + 'return ' + self.OutExpr(child[0]) + ';\n'
|
|
return self.dent() + 'return;\n'
|
|
if nt == 'DECL':
|
|
ret = self.dent() + node.t + ' ' + self.FindName(node)
|
|
if child:
|
|
if hasattr(child[0], 'orig') and (child[0].orig.nt != 'IDENT'
|
|
or child[0].orig.name
|
|
in self.symtab[child[0].orig.scope]):
|
|
ret += ' = ' + self.OutExpr(child[0].orig)
|
|
else:
|
|
ret += ' = ' + self.OutExpr(child[0])
|
|
return ret + ';\n'
|
|
if nt == ';':
|
|
return self.dent() + ';\n'
|
|
|
|
if nt in ('STDEF', '{}'):
|
|
ret = ''
|
|
if nt == 'STDEF':
|
|
if node.name == 'default':
|
|
ret = self.dent() + 'default\n'
|
|
else:
|
|
ret = self.dent() + 'state ' + self.FindName(node) + '\n'
|
|
|
|
ret += self.dent() + '{\n'
|
|
self.indentlevel += 1
|
|
for stmt in node.ch:
|
|
ret += self.OutCode(stmt)
|
|
self.indentlevel -= 1
|
|
return ret + self.dent() + '}\n'
|
|
|
|
if nt == 'FNDEF':
|
|
ret = self.dent()
|
|
if node.t is not None:
|
|
ret += node.t + ' '
|
|
ret += self.FindName(node) + '('
|
|
scope = node.pscope
|
|
ret += ', '.join(typ + ' ' + self.FindName(name, scope)
|
|
for typ, name in zip(node.ptypes, node.pnames))
|
|
return ret + ')\n' + self.OutCode(child[0])
|
|
|
|
if nt == 'EXPR':
|
|
return self.dent() + self.OutExpr(child[0]) + (
|
|
';\n' if not lslcommon.IsCalc else '')
|
|
|
|
if nt == 'LAMBDA':
|
|
return ''
|
|
|
|
assert False, "Internal error: node type not handled: " + nt # pragma: no cover
|
|
|
|
def output(self, treesymtab, options = ('optimize',
|
|
'optsigns','optfloats','warntabs')):
|
|
# Build a sorted list of dict entries
|
|
self.tree, self.symtab = treesymtab
|
|
|
|
# Grab options
|
|
self.optimize = 'optimize' in options
|
|
# These are optimization options that depend on the above:
|
|
self.optsigns = self.optimize and 'optsigns' in options
|
|
self.optfloats = self.optimize and 'optfloats' in options
|
|
self.foldconst = self.optimize and 'constfold' in options
|
|
|
|
self.warntabs = 'warntabs' in options
|
|
|
|
ret = ''
|
|
self.indent = ' '
|
|
self.indentlevel = 0
|
|
self.globalmode = False
|
|
self.listmode = False
|
|
for node in self.tree:
|
|
self.globalmode = node.nt == 'DECL'
|
|
ret += self.OutCode(node)
|
|
self.globalmode = False
|
|
|
|
return ret
|