def compactIf(node, thenPart, condition):
"""
Reduces the size of a if statement (without elsePart) using boolean operators instead of the typical keywords e.g.
"if(something)make()" is translated to "something&&make()" which is two characters shorter.
This however only works when the thenPart is only based on expressions and does not contain other statements.
"""
thenExpression = getattr(thenPart, "expression", None)
if not thenExpression:
# Empty semicolon statement => translate if into semicolon statement
node.remove(condition)
node.remove(node.thenPart)
node.append(condition, "expression")
node.type = "semicolon"
else:
# Has expression => Translate IF using a AND or OR operator
if condition.type == "not":
replacement = Node.Node(thenPart.tokenizer, "or")
condition = condition[0]
else:
replacement = Node.Node(thenPart.tokenizer, "and")
replacement.append(condition)
replacement.append(thenExpression)
thenPart.append(replacement, "expression")
fixParens(thenExpression)
fixParens(condition)
node.parent.replace(node, thenPart)
def __createSimpleAssignment(identifier, valueNode):
assignNode = Node.Node(None, "assign")
identNode = Node.Node(None, "identifier")
identNode.value = identifier
assignNode.append(identNode)
assignNode.append(valueNode)
return assignNode
def __rebuildAsAssignment(node, firstVarStatement):
"""Rebuilds the items of a var statement into a assignment list and moves declarations to the given var
statement."""
assignment = Node.Node(node.tokenizer, "semicolon")
assignmentList = Node.Node(node.tokenizer, "comma")
assignment.append(assignmentList, "expression")
# Casting to list() creates a copy during the process (keeps loop stable)
for child in list(node):
if hasattr(child, "name"):
# Cleanup initializer and move to assignment
if hasattr(child, "initializer"):
assign = __createSimpleAssignment(child.name,
child.initializer)
assignmentList.append(assign)
firstVarStatement.append(child)
else:
# JS 1.7 Destructing Expression
for identifier in child.names:
firstVarStatement.append(__createDeclaration(identifier.value))
if hasattr(child, "initializer"):
assign = __createMultiAssignment(child.names,
child.initializer)
assignmentList.append(assign)
node.remove(child)
# Patch parent node to contain assignment instead of declaration
if len(assignmentList) > 0:
node.parent.replace(node, assignment)
# Special process for "for-in" loops
# It is OK to be second because of assignments are not allowed at
# all in for-in loops and so the first if basically does nothing
# for these kind of statements.
elif getattr(node, "rel", None) == "iterator":
if hasattr(child, "name"):
node.parent.replace(node, __createIdentifier(child.name))
else:
# JS 1.7 Destructing Expressions
node.parent.replace(node, child.names)
# Edge case. Not yet found if this happen realistically
else:
if hasattr(node, "rel"):
Console.warn("Remove related node (%s) from parent: %s" %
(node.rel, node))
node.parent.remove(node)
# Minor post-cleanup. Remove useless comma statement when only one expression is the result
if len(assignmentList) == 1:
assignment.replace(assignmentList, assignmentList[0])
def __splitTemplate(value, valueParams):
"""
Split string into plus-expression(s)
- patchParam: string node containing the placeholders
- valueParams: list of params to inject
"""
# Convert list with nodes into Python dict
# [a, b, c] => {0:a, 1:b, 2:c}
mapper = {pos: value for pos, value in enumerate(valueParams)}
result = []
splits = __replacer.split(value)
if len(splits) == 1:
return None
pair = Node.Node(None, "plus")
for entry in splits:
if entry == "":
continue
if len(pair) == 2:
newPair = Node.Node(None, "plus")
newPair.append(pair)
pair = newPair
if __replacer.match(entry):
pos = int(entry[1]) - 1
# Items might be added multiple times. Copy to protect original.
try:
repl = mapper[pos]
except KeyError:
raise UserError("Invalid positional value: %s in %s" %
(entry, value))
copied = copy.deepcopy(mapper[pos])
if copied.type not in ("identifier", "call"):
copied.parenthesized = True
pair.append(copied)
else:
child = Node.Node(None, "string")
child.value = entry
pair.append(child)
return pair
def createHook(condition, thenPart, elsePart):
"""Creates a hook expression with the given then/else parts."""
hook = Node.Node(condition.tokenizer, "hook")
hook.append(condition, "condition")
hook.append(thenPart, "thenPart")
hook.append(elsePart, "elsePart")
return hook
def combineToCommaExpression(node):
"""
This method tries to combine a block with multiple statements into one semicolon statement with a comma expression
containing all expressions from the previous block. This only works when the block exclusively consists of
expressions as this do not work with other statements. Still this conversion reduces the size impact of many blocks
and leads to the removal of a lot of curly braces in the result code.
Example: {x++;y+=3} => x++,x+=3
"""
if node is None or node.type != "block":
return node
counter = 0
for child in node:
if child is None:
pass
elif child.type != "semicolon":
return node
else:
counter = counter + 1
if counter == 1:
return node
comma = Node.Node(node.tokenizer, "comma")
for child in list(node):
if child is None:
pass
# Ignore empty semicolons
if hasattr(child, "expression"):
comma.append(child.expression)
semicolon = Node.Node(node.tokenizer, "semicolon")
semicolon.append(comma, "expression")
parent = node.parent
parent.replace(node, semicolon)
return semicolon
def combineExpressions(condition, thenExpression, elseExpression):
"""Combines then and else expression using a hook statement."""
hook = createHook(condition, thenExpression, elseExpression)
semicolon = Node.Node(condition.tokenizer, "semicolon")
semicolon.append(hook, "expression")
fixParens(condition)
fixParens(thenExpression)
fixParens(elseExpression)
return semicolon
def reworkElse(node, elsePart):
"""
If an if ends with a return/throw we are able to inline the content of the else to the same parent as the if resides
into.
This method deals with all the nasty details of this operation.
"""
target = node.parent
targetIndex = target.index(node) + 1
# A workaround for compact if-else blocks
# We are a elsePart of the if where we want to move our
# content to. This cannot work. So we need to wrap ourself
# into a block and move the else statements to this newly
# established block
if not target.type in ("block", "script"):
newBlock = Node.Node(None, "block")
# Replace node with newly created block and put ourself into it
node.parent.replace(node, newBlock)
newBlock.append(node)
# Update the target and the index
target = newBlock
targetIndex = 1
if not target.type in ("block", "script"):
# print("No possible target found/created")
return elsePart
if elsePart.type == "block":
for child in reversed(elsePart):
target.insert(targetIndex, child)
# Remove else block from if statement
node.remove(elsePart)
else:
target.insert(targetIndex, elsePart)
return
def __combineVarStatements(node):
"""Top level method called to optimize a script node."""
if len(node.scope.declared) == 0:
return
firstVar = __findFirstVarStatement(node)
# Special case, when a node has variables, but no valid "var" block to hold them
# This happens in cases where there is a for-loop which contains a "var", but
# there are no other variable declarations anywhere. In this case we are not able
# to optimize the code further and just exit at this point
# Only size-saving when there are multiple for-in loops, but no other var statement or first
# "free" var declaration is after for-loops.
if not firstVar:
firstVar = Node.Node(None, "var")
node.insert(0, firstVar)
__patchVarStatements(node, firstVar)
__cleanFirst(firstVar)
# Remove unused "var"
if len(firstVar) == 0:
firstVar.parent.remove(firstVar)
else:
# When there is a classical for loop immediately after our
# first var statement, then we try to move the var declaration
# into there as a setup expression
firstVarParent = firstVar.parent
firstVarPos = firstVarParent.index(firstVar)
if len(firstVarParent) > firstVarPos + 1:
possibleForStatement = firstVarParent[firstVarPos + 1]
if possibleForStatement.type == "for" and not hasattr(
possibleForStatement, "setup"):
possibleForStatement.append(firstVar, "setup")
def __recurser(node, unused):
"""The cleanup part which always processes one scope and cleans up params and variable definitions which are
unused."""
retval = False
# Process children
if node.type != "function":
for child in node:
# None children are allowed sometimes e.g. during array_init like [1,2,,,7,8]
if child is not None:
if __recurser(child, unused):
retval = True
if node.type == "script" and hasattr(node, "parent"):
# Remove unused parameters
params = getattr(node.parent, "params", None)
if params:
# Start from back, as we can only remove params as long
# as there is not a required one after the current one
for identifier in reversed(params):
if identifier.value in unused:
Console.debug("Removing unused parameter '%s' in line %s",
identifier.value, identifier.line)
params.remove(identifier)
retval = True
else:
break
# Remove function names which are unused
if node.parent.functionForm == "expressed_form":
funcName = getattr(node.parent, "name", None)
if funcName is not None and funcName in unused:
Console.debug("Removing unused function name at line %s" %
node.line)
del node.parent.name
retval = True
elif node.type == "function":
# Remove full unused functions (when not in top-level scope)
if node.functionForm == "declared_form" and getattr(
node, "parent", None) and node.parent.type != "call":
funcName = getattr(node, "name", None)
if funcName is not None and funcName in unused:
Console.debug(
"Removing unused function declaration %s at line %s" %
(funcName, node.line))
node.parent.remove(node)
retval = True
elif node.type == "var":
for decl in reversed(node):
if getattr(decl, "name", None) in unused:
if hasattr(decl, "initializer"):
init = decl.initializer
if init.type in ("null", "this", "true", "false",
"identifier", "number", "string",
"regexp"):
Console.debug(
"Removing unused primitive variable %s at line %s"
% (decl.name, decl.line))
node.remove(decl)
retval = True
elif init.type == "function" and (
not hasattr(init, "name") or init.name in unused):
Console.debug(
"Removing unused function variable %s at line %s" %
(decl.name, decl.line))
node.remove(decl)
retval = True
# If we have only one child, we replace the whole var statement with just the init block
elif len(node) == 1:
semicolon = Node.Node(init.tokenizer, "semicolon")
semicolon.append(init, "expression")
# Protect non-expressions with parens
if init.type in ("array_init", "object_init"):
init.parenthesized = True
elif init.type == "call" and init[0].type == "function":
init[0].parenthesized = True
node.parent.replace(node, semicolon)
retval = True
# If we are the last declaration, move it out of node and append after var block
elif node[-1] == decl or node[0] == decl:
isFirst = node[0] == decl
node.remove(decl)
nodePos = node.parent.index(node)
semicolon = Node.Node(init.tokenizer, "semicolon")
semicolon.append(init, "expression")
# Protect non-expressions with parens
if init.type in ("array_init", "object_init"):
init.parenthesized = True
elif init.type == "call" and init[0].type == "function":
init[0].parenthesized = True
if isFirst:
node.parent.insert(nodePos, semicolon)
else:
node.parent.insert(nodePos + 1, semicolon)
retval = True
else:
Console.debug(
"Could not automatically remove unused variable %s at line %s without possible side-effects"
% (decl.name, decl.line))
else:
node.remove(decl)
retval = True
if len(node) == 0:
Console.debug("Removing empty 'var' block at line %s" % node.line)
node.parent.remove(node)
return retval
def __recurser(node, table):
counter = 0
# Process children
for child in list(node):
if child is not None:
counter += __recurser(child, table)
# Process all method calls
if node.type == "call":
funcName = None
funcNameNode = None
# Uses global translation method (not typical)
if node[0].type == "identifier":
funcNameNode = node[0]
# Uses namespaced translation method.
# Typically core.locale.Translation.tr() or Translation.tr()
elif node[0].type == "dot" and node[0][1].type == "identifier":
funcNameNode = node[0][1]
# Gettext methods only at the moment
funcName = funcNameNode and funcNameNode.value
if funcName in translationFunctions:
Console.debug("Found translation method %s in %s", funcName,
node.line)
Console.indent()
params = node[1]
# Remove marktr() calls
if funcName == "marktr":
node.parent.remove(node)
# Verify param types
elif params[0].type is not "string":
# maybe something marktr() relevant being used, in this case we need to keep the call and inline the data
pass
# Error handling
elif (funcName == "trn"
or funcName == "trc") and params[1].type != "string":
Console.warn(
"Expecting translation string to be type string: %s at line %s"
% (params[1].type, params[1].line))
# Signature tr(msg, arg1, ...)
elif funcName == "tr":
key = params[0].value
if key in table:
params[0].value = table[key]
counter += 1
if len(params) == 1:
node.parent.replace(node, params[0])
else:
replacement = __splitTemplate(params[0].value, params[1:])
if replacement:
node.parent.replace(node, replacement)
# Signature trc(context, msg, arg1, ...)
elif funcName == "trc":
key = "%s[C:%s]" % (params[1].value, params[0].value)
if key in table:
params[1].value = table[key]
counter += 1
if len(params) == 2:
node.parent.replace(node, params[1])
else:
replacement = __splitTemplate(params[1].value, params[2:])
if replacement:
node.parent.replace(node, replacement)
# Signature trn(msgSingular, msgPlural, int, arg1, ...)
elif funcName == "trn":
key = "%s[N:%s]" % (params[0].value, params[1].value)
if key not in table:
Console.outdent()
return counter
counter += 1
# Use optimized trnc() method instead of trn()
funcNameNode.value = "trnc"
# Remove first two string parameters
params.remove(params[0])
params.remove(params[0])
# Inject new object into params
container = Node.Node(None, "object_init")
params.insert(0, container)
# Create new construction with all properties generated from the translation table
for plural in table[key]:
pluralEntry = Node.Node(None, "property_init")
pluralEntryIdentifier = Node.Node(None, "identifier")
pluralEntryIdentifier.value = plural
pluralEntryValue = Node.Node(None, "string")
pluralEntryValue.value = table[key][plural]
pluralEntry.append(pluralEntryIdentifier)
pluralEntry.append(pluralEntryValue)
container.append(pluralEntry)
# Replace strings with plus operations to omit complex client side string operation
if len(params) > 2:
for pluralEntry in container:
replacement = __splitTemplate(pluralEntry[1].value,
params[2:])
if replacement:
pluralEntry.replace(pluralEntry[1], replacement)
# When all variables have been patched in all string with placeholder
# we are able to remove the whole list of placeholder values afterwards
while len(params) > 2:
params.pop()
Console.outdent()
return counter
def createReturn(value):
"""Creates a return statement with the given value."""
ret = Node.Node(value.tokenizer, "return")
ret.append(value, "value")
return ret
def __optimize(node, compressor):
# Process from inside to outside
# on a copy of the node to prevent it from forgetting children when structure is modified
for child in list(node):
# None children are allowed sometimes e.g. during array_init like [1,2,,,7,8]
if child is not None:
__optimize(child, compressor)
# Cleans up empty semicolon statements (or pseudo-empty)
if node.type == "semicolon" and node.parent.type in ("block", "script"):
expr = getattr(node, "expression", None)
if not expr or expr.type in ("null", "this", "true", "false",
"identifier", "number", "string",
"regexp"):
# Keep scrict mode hints
if expr and expr.type is "string" and expr.value == "use strict":
pass
else:
if expr is not None:
Console.debug(
"Remove empty statement at line %s of type: %s",
expr.line, expr.type)
node.parent.remove(node)
return
# Remove unneeded parens
if getattr(node, "parenthesized", False):
cleanParens(node)
# Pre-compute numeric expressions where it makes sense
if node.type in (
"plus", "minus", "mul", "div",
"mod") and node[0].type == "number" and node[1].type == "number":
firstNumber = node[0]
secondNumber = node[1]
operator = node.type
# Only do for real numeric values and not for protected strings (float differences between Python and JS)
if isinstance(firstNumber.value, str) or isinstance(
secondNumber.value, str):
pass
elif operator == "plus":
Console.debug("Precompute numeric %s operation at line: %s",
operator, node.line)
firstNumber.value += secondNumber.value
node.parent.replace(node, firstNumber)
elif operator == "minus":
Console.debug("Precompute numeric %s operation at line: %s",
operator, node.line)
firstNumber.value -= secondNumber.value
node.parent.replace(node, firstNumber)
else:
if operator == "mul":
result = firstNumber.value * secondNumber.value
elif operator == "div" and secondNumber.value is not 0:
result = firstNumber.value / secondNumber.value
elif operator == "mod":
result = firstNumber.value % secondNumber.value
else:
result = None
if result is not None and len(str(result)) < len(
compressor.compress(node)):
Console.debug("Precompute numeric %s operation at line: %s",
operator, node.line)
firstNumber.value = result
node.parent.replace(node, firstNumber)
# Pre-combine strings (even supports mixed string + number concats)
elif node.type == "plus" and node[0].type in (
"number", "string") and node[1].type in ("number", "string"):
Console.debug("Joining strings at line: %s", node.line)
node[0].value = "%s%s" % (node[0].value, node[1].value)
node[0].type = "string"
node.parent.replace(node, node[0])
# Pre-combine last with last (special case e.g.: somevar + "hello" + "world")
elif node.type == "plus" and node[0].type == "plus" and node[0][
1].type in ("number", "string") and node[1].type in (
"number", "string") and node[0][1].type == node[1].type:
node[1].value = "%s%s" % (node[0][1].value, node[1].value)
node[1].type = "string"
node.replace(node[0], node[0][0])
# Unwrap blocks
if node.type == "block":
if node.parent.type in ("try", "catch", "finally"):
pass
elif len(node) == 0:
Console.debug("Replace empty block with semicolon at line: %s",
node.line)
repl = Node.Node(node.tokenizer, "semicolon")
node.parent.replace(node, repl)
node = repl
elif len(node) == 1:
if node.parent.type == "if" and node.rel == "thenPart" and hasattr(
node.parent, "elsePart") and containsIf(node):
# if with else where the thenBlock contains another if
pass
elif node.parent.type == "if" and node.rel == "thenPart" and containsIfElse(
node):
# if without else where the thenBlock contains a if-else
pass
elif node.parent.type in ("case", "default"):
# virtual blocks inside case/default statements
pass
else:
# debug("Removing block for single statement at line %s", node.line)
node.parent.replace(node, node[0])
node = node[0]
else:
node = combineToCommaExpression(node)
# Remove "empty" semicolons which are inside a block/script parent
if node.type == "semicolon":
if not hasattr(node, "expression"):
if node.parent.type in ("block", "script"):
Console.debug("Remove empty semicolon expression at line: %s",
node.line)
node.parent.remove(node)
elif node.parent.type == "if":
rel = getattr(node, "rel", None)
if rel == "elsePart":
Console.debug("Remove empty else part at line: %s",
node.line)
node.parent.remove(node)
# Process all if-statements
if node.type == "if":
condition = node.condition
thenPart = node.thenPart
elsePart = getattr(node, "elsePart", None)
# Optimize for empty thenPart if elsePart is available
if thenPart.type == "semicolon" and not hasattr(
thenPart, "expression") and elsePart:
if condition.type == "not":
node.replace(condition, condition[0])
condition = condition[0]
else:
repl = Node.Node(None, "not")
node.replace(condition, repl)
repl.append(condition)
fixParens(condition)
condition = repl
node.replace(thenPart, elsePart)
thenPart = elsePart
elsePart = None
# Optimize using hook operator
if elsePart and thenPart.type == "return" and elsePart.type == "return" and hasattr(
thenPart, "value") and hasattr(elsePart, "value"):
# Combine return statement
replacement = createReturn(
createHook(condition, thenPart.value, elsePart.value))
node.parent.replace(node, replacement)
return
# Check whether if-part ends with a return statement. Then
# We do not need a else statement here and just can wrap the whole content
# of the else block inside the parent
if elsePart and endsWithReturnOrThrow(thenPart):
reworkElse(node, elsePart)
elsePart = None
# Optimize using "AND" or "OR" operators
# Combine multiple semicolon statements into one semicolon statement using an "comma" expression
thenPart = combineToCommaExpression(thenPart)
elsePart = combineToCommaExpression(elsePart)
# Optimize remaining if or if-else constructs
if elsePart:
mergeParts(node, thenPart, elsePart, condition, compressor)
elif thenPart.type == "semicolon":
compactIf(node, thenPart, condition)
def __createIdentifier(value):
identifier = Node.Node(None, "identifier")
identifier.value = value
return identifier
def __createDeclaration(name):
declNode = Node.Node(None, "declaration")
declNode.name = name
declNode.readOnly = False
return declNode
def __createMultiAssignment(names, valueNode):
assignNode = Node.Node(None, "assign")
assignNode.append(names)
assignNode.append(valueNode)
return assignNode