Added SheerkaComparisonManager

This commit is contained in:
2020-05-17 20:19:26 +02:00
parent 56e0a9d338
commit 08e3086820
29 changed files with 586 additions and 148 deletions
+247
View File
@@ -0,0 +1,247 @@
import core.utils
from core.builtin_concepts import BuiltinConcepts, ReturnValueConcept
NO_MATCH = "** No Match **"
class SheerkaExecute:
"""
Manage the execution of a process flow
"""
def __init__(self, sheerka):
self.sheerka = sheerka
def call_parsers(self, execution_context, return_values):
# return_values must be a list
if not isinstance(return_values, list):
return_values = [return_values]
# first make the distinguish between what is for the parsers and what is not
result = []
to_process = []
for r in return_values:
if not r.status or not self.sheerka.isinstance(r.body, BuiltinConcepts.USER_INPUT):
result.append(r)
else:
to_process.append(r)
if not to_process:
return result
# keep track of the originals user inputs, as they need to be removed at the end
user_inputs = to_process[:]
# group the parsers by priorities
instantiated_parsers = [parser(sheerka=self.sheerka) for parser in self.sheerka.parsers.values()]
instantiated_parsers = self.preprocess(execution_context, instantiated_parsers)
grouped_parsers = {}
for parser in [p for p in instantiated_parsers if p.enabled]:
grouped_parsers.setdefault(parser.priority, []).append(parser)
sorted_priorities = sorted(grouped_parsers.keys(), reverse=True)
stop_processing = False
for priority in sorted_priorities:
inputs_for_this_group = to_process[:]
for parser in grouped_parsers[priority]:
for return_value in inputs_for_this_group:
to_parse = return_value.body.body \
if self.sheerka.isinstance(return_value.body, BuiltinConcepts.USER_INPUT) \
else return_value.body
# if self.sheerka.log.isEnabledFor(logging.DEBUG):
# debug_text = "'" + to_parse + "'" if isinstance(to_parse, str) \
# else "'" + BaseParser.get_text_from_tokens(to_parse) + "' as tokens"
# execution_context.log(f"Parsing {debug_text}")
with execution_context.push(desc=f"Parsing using {parser.name}",
logger=parser.verbose_log) as sub_context:
sub_context.add_inputs(to_parse=to_parse)
res = parser.parse(sub_context, to_parse)
if res is not None:
if hasattr(res, "__iter__"):
for r in res:
if r is None:
continue
r.parents = [return_value]
result.append(r)
if self.sheerka.isinstance(r.body, BuiltinConcepts.PARSER_RESULT):
# if a ParserResultConcept is returned, it will be used by the parsers
# of the following groups
to_process.append(r)
if r.status:
stop_processing = True
else:
res.parents = [return_value]
result.append(res)
if self.sheerka.isinstance(res.body, BuiltinConcepts.PARSER_RESULT):
# if a ParserResultConcept is returned, it will be used by the parsers
# of the following groups
to_process.append(res)
if res.status:
stop_processing = True
sub_context.add_values(return_values=res)
if stop_processing:
break # Do not try the other priorities if a match is found
result = core.utils.remove_list_from_list(result, user_inputs)
return result
def call_evaluators(self, execution_context, return_values, process_step):
# return_values must be a list
if not isinstance(return_values, list):
return_values = [return_values]
# group the evaluators by priority and sort them
# The first one to be applied will be the one with the highest priority
grouped_evaluators = {}
instantiated_evaluators = [e_class() for e_class in self.sheerka.evaluators]
# pre-process evaluators if needed
instantiated_evaluators = self.preprocess(execution_context, instantiated_evaluators)
for evaluator in [e for e in instantiated_evaluators if e.enabled and process_step in e.steps]:
grouped_evaluators.setdefault(evaluator.priority, []).append(evaluator)
# order the groups by priority, the higher first
sorted_priorities = sorted(grouped_evaluators.keys(), reverse=True)
# process
iteration = 0
while True:
with execution_context.push(desc=f"iteration #{iteration}", iteration=iteration) as iteration_context:
simple_digest = return_values[:]
iteration_context.add_inputs(return_values=simple_digest)
for priority in sorted_priorities:
original_items = return_values[:]
evaluated_items = []
to_delete = []
for evaluator in grouped_evaluators[priority]:
evaluator = self.preprocess(execution_context, evaluator.__class__()) # fresh copy
sub_context_desc = f"Evaluating using {evaluator.name} ({priority=})"
with iteration_context.push(desc=sub_context_desc, logger=evaluator.verbose_log) as sub_context:
sub_context.add_inputs(return_values=original_items)
# process evaluators that work on one simple return value at the time
from evaluators.BaseEvaluator import OneReturnValueEvaluator
if isinstance(evaluator, OneReturnValueEvaluator):
debug_result = []
for item in original_items:
if evaluator.matches(sub_context, item):
# init the evaluator is possible
if hasattr(evaluator, "init_evaluator") and not evaluator.is_initialized:
evaluator.init_evaluator(sub_context, original_items)
result = evaluator.eval(sub_context, item)
if result is None:
debug_result.append({"input": item, "return_value": None})
continue
to_delete.append(item)
if isinstance(result, list):
evaluated_items.extend(result)
elif isinstance(result, ReturnValueConcept):
evaluated_items.append(result)
else:
error = self.sheerka.new(BuiltinConcepts.INVALID_RETURN_VALUE, body=result,
evaluator=evaluator)
result = self.sheerka.ret("sheerka.process", False, error, parents=[item])
evaluated_items.append(result)
debug_result.append({"input": item, "return_value": result})
else:
debug_result.append({"input": item, "return_value": NO_MATCH})
sub_context.add_values(return_values=debug_result)
# process evaluators that work on all return values
else:
if evaluator.matches(sub_context, original_items):
results = evaluator.eval(sub_context, original_items)
if results is None:
continue
if not isinstance(results, list):
results = [results]
for result in results:
evaluated_items.append(result)
to_delete.extend(result.parents)
sub_context.add_values(return_values=results)
else:
sub_context.add_values(return_values=NO_MATCH)
return_values = evaluated_items
return_values.extend([item for item in original_items if item not in to_delete])
iteration_context.add_values(return_values=return_values[:])
# have we done something ?
to_compare = return_values[:]
if simple_digest == to_compare:
break
# inc the iteration and continue
iteration += 1
return return_values
def execute(self, execution_context, return_values, execution_steps):
"""
Executes process for all initial contexts
:param execution_context:
:param return_values:
:param execution_steps:
:return:
"""
for step in execution_steps:
copy = return_values[:] if hasattr(return_values, "__iter__") else [return_values]
with execution_context.push(step=step, iteration=0, desc=f"{step=}") as sub_context:
if step == BuiltinConcepts.PARSING:
return_values = self.call_parsers(sub_context, return_values)
else:
return_values = self.call_evaluators(sub_context, return_values, step)
if copy != return_values:
sub_context.log_result(return_values)
sub_context.add_values(return_values=return_values)
return return_values
def preprocess(self, context, parsers_or_evaluators):
if not context.preprocess:
return parsers_or_evaluators
if not hasattr(parsers_or_evaluators, "__iter__"):
single_one = True
parsers_or_evaluators = [parsers_or_evaluators]
else:
single_one = False
for preprocess in context.preprocess:
for e in parsers_or_evaluators:
if self.matches(e.name, preprocess.get_value("name")):
for var_name in preprocess.values:
if var_name == "name":
continue
if hasattr(e, var_name):
setattr(e, var_name, preprocess.get_value(var_name))
return parsers_or_evaluators[0] if single_one else parsers_or_evaluators
@staticmethod
def matches(parser_or_evaluator_name, preprocessor_name):
if preprocessor_name.endswith("*"):
return parser_or_evaluator_name.startswith(preprocessor_name[:-1])
else:
return parser_or_evaluator_name == preprocessor_name