def test_nested_findAllNested(self):
s = Scope.empty()
s.add(self.obj1)
ns = s.nestedScope()
ns.add(self.obj2)
ns.add(self.obj3)
it = ns.findAllNested()
r = [obj for obj in it]
self.assertEqual(r, [self.obj3, self.obj2])
def test_call_is_last_in(self):
s = Scope.empty()
e1 = TestEntry('plod', 3)
s.add(e1)
e2 = TestEntry('plod', 3)
s.add(e2)
e3 = TestEntry('plod', 3)
s.add(e3)
r = s('plod')
self.assertEqual(r, e3)
def test_findAllByName(self):
s = Scope.empty()
e1 = TestEntry('plod', 3)
s.add(e1)
e2 = TestEntry('plod', 3)
s.add(e2)
e3 = TestEntry('plod', 3)
s.add(e3)
it = s.findAllByName('plod')
r = [obj for obj in it]
self.assertEqual(r, [e3, e2, e1])
def __init__(self, tokenIt, builderAPI):
'''
builderAPI
'''
self.b = Builder()
self.instructionStack = []
self.instructionsStoreTrigger = False
# EnvStd is builtin symbol definitions from BuilderAPI
#x
self.envStd = builderAPI
self.builderAPI = builderAPI
builderFuncSymbols = [ SymbolBuiltinFunc(funcName, getattr(builderAPI, funcName), NoType) for funcName in builderAPI.funcNameToArgsType.keys()]
self.scopeStd = Scope(builderFuncSymbols)
# These need explaining.
# This is a link auto-wired into the API so the API can refer
# back to this class.
# Really the API is a specialism of this class, and should
# inherit it. But then it is not an API, and inherits plenty
# of methods and attributes that may get awkward.
# So the API is composed, and so cleanly encapsulated.
# Main issue with that is that the API benefits from access to
# this class. It will recieve further errors from construction
# classes, and sometimes needs to signal the overall handlers
# here to work e.g. make a new environment.
# If the API was interited, it could do that, but composed, no.
# For a while the solution was to send data back with
# structures like Options and Eithers. But this got messy, as
# there is a lot of it. And the API knows what it wants to do,
# no need for ifs...
# So the API is now given a hard link back here to do it's
# tinkering. If there is one problen, you can see a link loop
# now exists, the API can call into the Compiler affecting the
# API... However, the concerns and information are clearly
# outlined, this should never occur.
#x
self.envStd.compiler = self
self.builderAPI.compiler = self
self.funcNameToArgsType = builderAPI.funcNameToArgsType
# EnvClosure holds symbiol definitions local to patches of
# code, such as codeblocks for whiles etc.
# As such, it can be stacked.
#x
self.envClosure = []
self.scopeStack = []
# Used for the occasional section vars.
#? Policy undecided
#x
#self.envGlobal = {}
self.scopeGlobal = Scope.empty()
# Closure data is not part of the env group at all.
# Rubble does not have bracketing, and does not enable
# bracketing, prefering to mark start and end locations with
# sitandalone functions. So we need to keep track of those.
# Of course, environments will be enabled within these blocks.
self.closureData = []
super().__init__(tokenIt)
class Compiler(Syntaxer):
def __init__(self, tokenIt, builderAPI):
'''
builderAPI
'''
self.b = Builder()
self.instructionStack = []
self.instructionsStoreTrigger = False
# EnvStd is builtin symbol definitions from BuilderAPI
#x
self.envStd = builderAPI
self.builderAPI = builderAPI
builderFuncSymbols = [ SymbolBuiltinFunc(funcName, getattr(builderAPI, funcName), NoType) for funcName in builderAPI.funcNameToArgsType.keys()]
self.scopeStd = Scope(builderFuncSymbols)
# These need explaining.
# This is a link auto-wired into the API so the API can refer
# back to this class.
# Really the API is a specialism of this class, and should
# inherit it. But then it is not an API, and inherits plenty
# of methods and attributes that may get awkward.
# So the API is composed, and so cleanly encapsulated.
# Main issue with that is that the API benefits from access to
# this class. It will recieve further errors from construction
# classes, and sometimes needs to signal the overall handlers
# here to work e.g. make a new environment.
# If the API was interited, it could do that, but composed, no.
# For a while the solution was to send data back with
# structures like Options and Eithers. But this got messy, as
# there is a lot of it. And the API knows what it wants to do,
# no need for ifs...
# So the API is now given a hard link back here to do it's
# tinkering. If there is one problen, you can see a link loop
# now exists, the API can call into the Compiler affecting the
# API... However, the concerns and information are clearly
# outlined, this should never occur.
#x
self.envStd.compiler = self
self.builderAPI.compiler = self
self.funcNameToArgsType = builderAPI.funcNameToArgsType
# EnvClosure holds symbiol definitions local to patches of
# code, such as codeblocks for whiles etc.
# As such, it can be stacked.
#x
self.envClosure = []
self.scopeStack = []
# Used for the occasional section vars.
#? Policy undecided
#x
#self.envGlobal = {}
self.scopeGlobal = Scope.empty()
# Closure data is not part of the env group at all.
# Rubble does not have bracketing, and does not enable
# bracketing, prefering to mark start and end locations with
# sitandalone functions. So we need to keep track of those.
# Of course, environments will be enabled within these blocks.
self.closureData = []
super().__init__(tokenIt)
def instructionsStore(self):
'''
Set a new builder as the current builder,
For local manipilation of built code, such as repetitions
or multiple inserts.
'''
self.instructionStack.append([])
self.instructionsStoreTrigger = True
def instructionsGet(self):
'''
Return the current builder as a result.
This will revert the current builder to the previous builder.
'''
#assert (len(self.instructionStack) > 0), "This error should not occur!!! On builderResult instructionStack is empty."
return self.instructionStack.pop(-1)
def instructionsPlay(self, instructions):
'''
Play stored instructions
'''
#! currently works on new env but current builder, which is wrong
# what about embedded loops, huh?
for ins in instructions:
# Aye, unPythonic
pos = ins[0]
#posArgs = ins[1]
name = ins[1]
args = ins[2]
self.exprCB(pos, name, args)
def _stringTypeNamesMk(self, argTests):
'''
print a list of human-readable typenames
'''
# Used on argument signatures to tidy error reports
#return "[" + ", ".join([tpe.__name__ for tpe in typeList]) + "]"
return "[" + ", ".join([argTest.description for argTest in argTests]) + "]"
def argsCheck(self, parsedArgs, argTests):
'''
Check args against signature.
The signature is a list of tests. This function first checks
that the length of the arg list matches that of the given
signature. Then it steps each test and argument, applying the
test.
argTests
a list of callable functions to test an arg
'''
args = parsedArgs.value
argsPos = parsedArgs.position
if (len(args) > len(argTests)):
msg = "Too many args. expected:{}".format(
self._stringTypeNamesMk(argTests),
)
self.errorWithPos(argsPos, msg)
if (len(args) < len(argTests)):
msg = "Not enough args. expected:{}".format(
self._stringTypeNamesMk(argTests),
)
self.errorWithPos(argsPos, msg)
i = 0
for argTest, arg in zip(argTests, args):
if (not(argTest(arg))):
#if (not(argTest(arg))):
#? good for debugging, but no general purpose
print(f'...failing args check. test:{argTest}. arg:{arg}')
msg = "Arg type not match signature. expected:{}".format(
argTest.description,
)
self.errorWithPos(arg.position, msg)
i += 1
def argsError(self, posArgs, either):
if (either.status == Option.ERROR):
self.errorWithPos(posArgs, either.msg)
if (either.status == Option.WARNING):
self.warningWithPos(posArgs, either.msg)
if (either.status == Option.INFO):
self.info(either.msg)
## ClosureData
def closureDataPush(self, data):
return self.closureData.append(data)
def closureDataPeek(self):
if (not(self.closureData)):
self.error('Peek codeblock data when no block open.')
return self.closureData[-1]
def closureDataPop(self):
if (not(self.closureData)):
self.error('Close a codeblock when no block open?')
return self.closureData.pop()
## environments
#! it would maybe be faster to make this mutable
# Then put an env on tracking new vars
# Then delete the new vars at the end.
# This would be a lot faster, actually, rather than the copy()
#x
# def envAddClosure(self):
# newEnv = {}
# if (self.envClosure):
# # Copy all lower symbols into top layer
# # Probably slow to copy, but easy to
# # find and search for symbols.
# # Coppy because don't want adaptions seeping back to wider
# # blocks.
# newEnv = dict(self.envClosure[-1])
# self.envClosure.append(newEnv)
def scopeStackPush(self):
scope = Scope.empty()
# link up symbols from the super-scope
if (self.scopeStack):
scope.addScope(self.scopeStack[-1])
self.scopeStack.append(scope)
#x
# def envDelClosure(self):
# assert(self.envClosure), "Celete non-existant envCloseure."
# del(self.envClosure[-1])
def scopeStackPop(self):
assert(self.scopeStack), "Celete non-existant scope."
del(self.scopeStack[-1])
#x
# def symbolSet(self, protoSymbol, value):
# '''
# Register a symbol to the current environment.
# '''
# assert(self.envClosure), "Symbol offered, but no envClosure. protoSymbol:{}".format(
# protoSymbol
# )
# #print('setting: ' + protoSymbol)
# self.envClosure[-1][protoSymbol] = value
def symSet(self, sym):
'''
Register a symbol to the current environment.
'''
assert(self.scopeStack), "Symbol offered, but no scopeStack. protoSymbol:{}".format(
protoSymbol
)
#print('setting: ' + protoSymbol)
self.scopeStack[-1].add(sym)
#x
# def symbolUpdateType(self, name, tpe):
# '''
# A brute mutation of a already registered type
# Used for genVars
# '''
# self.envClosure[-1][name].tpe = tpe
# May yet be used for clutch iteration on genVars?
def symUpdateType(self, name, tpe):
'''
A brute mutation of a already registered type
Used for genVars
'''
self.scopeStack[-1](name).tpe = tpe
#x
# def symbolUpdateLoc(self, name, loc):
# '''
# Unregister a symbol from the current environment.
# Despite environment cleanup, this is required. It is needed for
# the circumstance where a var is overwritten, so presumably
# has no further purpose.
# '''
# self.envClosure[-1][name].loc = loc
#x done by pokes in UpdateLocationBuilder. Anything else necessary?
def symUpdateLoc(self, name, loc):
'''
Unregister a symbol from the current environment.
Despite environment cleanup, this is required. It is needed for
the circumstance where a var is overwritten, so presumably
has no further purpose.
'''
self.scopeStack[-1](name).loc = loc
#x is this now used? We kill and create scopes even for small items
# def symbolDelete(self, protoSymbol, value):
# '''
# A brute mutation of a already registered type
# '''
# #? Used for genVars
# assert(self.envClosure), "Symbol offered, but no envClosure. protoSymbol:{}".format(
# protoSymbol
# )
# #print('setting: ' + protoSymbol)
# del(self.envClosure[-1][protoSymbol])
#x
# def envPrint(self):
# #! oh so yes! Sadly, the same bad defaults noted in exprCB()
# #print(str(self.envClosure[-1]))
# print('env:')
# for k,v in self.envClosure[-1].items():
# print(k + ': ' + str(v))
def scopeGlobalPrint(self):
print('env Global:')
for e in self.scopeGlobal.toList():
print(f" {e.name}: {e.tpe}")
def scopePrint(self):
#! oh so yes! Sadly, the same bad defaults noted in exprCB()
#print(str(self.scopeStack[-1]))
print('env Usr:'******'env Builtin:')
#? Be nice to print args sigs too?
for e in self.scopeStd.toList():
print(' ' + e.name)
#x
# def symbolSetGlobal(self, protoSymbol, value):
# # Used in RO on builder
# #? Do a value test, or not
# print('Global setting: ' + protoSymbol)
# self.envGlobal[protoSymbol] = value
def symbolSetGlobal(self, symVar):
# Used in RO on builder
#? Do a value test, or not
#print('Global setting: ' + symVar.name)
self.scopeGlobal.add(symVar)
#x
# def findIdentifier(self, pos, sym):
# # envStd overrides all
# if (sym in self.envStd):
# return self.envStd[sym]
# # All envClosures have wider closures loaded too
# if (self.envClosure and (sym in self.envClosure[-1])):
# return self.envClosure[-1][sym]
# # last shot, globals
# #if (sym in self.envGlobal):
# # return self.envGlobal[sym]
# # last shot, globals
# #print(sym)
# symMaybe = self.scopeGlobal(sym)
# if (symMaybe):
# return symMaybe
# self.envPrint()
# msg = "Symbol requested but not found in scope. symbol '{}'".format(
# sym
# )
# self.errorWithPos(pos, msg)
def symbolBuiltinFind(self, pos, symName):
symMaybe = self.scopeStd(symName)
if (symMaybe):
return symMaybe
self.scopePrint()
msg = "Symbol requested but not found in scope. symName '{}'".format(
symName
)
self.errorWithPos(pos, msg)
def symbolUsrFind(self, pos, symName):
#? Used in two places, and I'm thinking it shouldn't be?
# - To find builtin funcnames in the dispatcher below
# - in the syntaxer, to type arguments
# scopeStd overrides all
#symMaybe = self.scopeStd(symName)
#if (symMaybe):
# return symMaybe
# All scopeStacks have wider scopes loaded too
if (self.scopeStack):
symMaybe = self.scopeStack[-1](symName)
if (symMaybe):
return symMaybe
# last shot, globals
symMaybe = self.scopeGlobal(symName)
if (symMaybe):
return symMaybe
#print('findIdentifier')
self.scopePrint()
msg = "Symbol requested but not found in scope. symName '{}'".format(
symName
)
self.errorWithPos(pos, msg)
## Syntaxer callbacks
def commentCB(self, text):
print('Compiler comment with "' + text)
def exprCB(self, pos, name, args):
#! such a useful print---enhance and be part of a debug?
# printing a list always REPRs, think this is due a change in
# 3.7 I prefer the string marks to a gross repr. So the
# list comprehension.
print('Compiler expr {}({!s})'.format(
name,
#[str(a) for a in args]
[str(a) for a in args.value]
))
if (self.instructionsStoreTrigger):
#! need something more general than this hardcode
# I do want to control this, because it will lead to incomprehensible arrors,
# so catch whenever, hardcoded? e.g.
# if (name in storeCloseFuncs):
if (
name == 'forEachUnrolledEnd'
or name == 'whenEnd'
):
self.instructionsStoreTrigger = False
# play the end instruction
self.exprCB(pos, name, args)
else:
self.instructionStack[-1].append((pos, name, args,))
else:
#x
#func = self.findIdentifier(pos, name)
#! No, this dispatcher only needs to search in stdSymbols
# and funcDefs, I think?
#self.scopeStdPrint()
func = self.symbolBuiltinFind(pos, name)
# stacked data protection
#? assert?
if ((name == "funcEnd" or name == "funcMainEnd") and len(self.closureData) > 0):
msg = "End of func with unclosed instructions. unused instruction args:{}".format(
self.closureData,
)
self.errorWithPos(pos, msg)
# Test args for type and/or count
#i these tests acceept the The(position) wrap
#! should use at all if not registered? i.e. not need this if?
if name in self.funcNameToArgsType:
self.argsCheck(args, self.funcNameToArgsType[name])
# Wafow--now can do a simple call
#i AFASIK, no need here to send the full arg position. No
# call will use it. So strip, because it removes a lot of
# repetitive code
msgOption = func.data(self.b, args.value)
# If an message came from the API, its an integrity
# error from the args (not a function issue)
# Outright throws are throws. Throws to be caught are in custom
# exceptions of this code.
#! why extract the arg?
#? do need this at all
self.argsError(args.position, msgOption)
def result(self):
assert (len(self.instructionStack) == 0), "This error should not occur!!! On Compiler result, instructionStack is not empty. len:{}".format(len(self.instructionStack))
return self.b
def test_nested_depth(self):
s = Scope.empty()
s.add(self.obj1)
ns = s.nestedScope()
self.assertEqual(ns.depth, 1)
def test_call_fail(self):
s = Scope.empty()
s.add(self.obj1)
r = s('slowly')
self.assertEqual(r, None)
def test_call(self):
s = Scope.empty()
s.add(self.obj1)
r = s('plod')
self.assertEqual(r, self.obj1)
def test_toList(self):
s = Scope.empty()
s.add(self.obj1)
s.add(self.obj2)
self.assertEqual(s.toList(), [self.obj2, self.obj1 ])
def test_size_multi(self):
s = Scope.empty()
s.add(self.obj1)
s.add(self.obj2)
self.assertEqual(s.size(), 2)
def test_size_non_empty(self):
s = Scope.empty()
s.add(self.obj1)
self.assertEqual(s.size(), 1)
def test_add_fail(self):
s = Scope.empty()
with self.assertRaises(AssertionError):
s.add(5)
def test_add(self):
s = Scope.empty()
s.add(self.obj1)
self.assertEqual(s.depth, 0)
def test_size(self):
s = Scope.empty()
self.assertEqual(s.size(), 0)
def test_empty(self):
s = Scope.empty()
self.assertEqual(s.depth, 0)
def test_nested_inherits(self):
s = Scope.empty()
s.add(self.obj1)
ns = s.nestedScope()
r = s('plod')
self.assertEqual(r, self.obj1)
