Merge pull request #275 from hartym/better_errors

Proposed changes about errors and casts.

Makefile

@@ -1,4 +1,4 @@
-# Generated by Medikit 0.6.3 on 2018-06-11.
+# Generated by Medikit 0.6.3 on 2018-07-22.
 # All changes will be overriden.
 # Edit Projectfile and run “make update” (or “medikit update”) to regenerate.
 

bonobo/config/configurables.py

@@ -72,7 +72,7 @@ class ConfigurableMeta(type):
 
 try:
     import _functools
-except:
+except ImportError:
     import functools
 
     PartiallyConfigured = functools.partial
@@ -141,15 +141,14 @@ class Configurable(metaclass=ConfigurableMeta):
                 break  # option orders make all positional options first, job done.
 
             if not isoption(getattr(cls, name)):
-                missing.remove(name)
+                missing.discard(name)
                 continue
 
             if len(args) <= position:
                 break  # no more positional arguments given.
 
             position += 1
-            if name in missing:
-                missing.remove(name)
+            missing.discard(name)
 
         # complain if there is more options than possible.
         extraneous = set(kwargs.keys()) - (set(next(zip(*options))) if len(options) else set())

bonobo/config/services.py

@@ -1,5 +1,3 @@
-import inspect
-import pprint
 import re
 import threading
 import types

bonobo/examples/datasets/__main__.py

@@ -51,7 +51,7 @@ if __name__ == '__main__':
                         s3.head_object(
                             Bucket='bonobo-examples', Key=s3_path
                         )
-                    except:
+                    except Exception:
                         s3.upload_file(
                             local_path,
                             'bonobo-examples',

bonobo/execution/contexts/base.py

@@ -1,12 +1,10 @@
 import logging
 import sys
 from contextlib import contextmanager
-from logging import ERROR
-
-from mondrian import term
 
 from bonobo.util import deprecated
 from bonobo.util.objects import Wrapper, get_name
+from mondrian import term
 
 
 @contextmanager
@@ -14,7 +12,7 @@ def recoverable(error_handler):
     try:
         yield
     except Exception as exc:  # pylint: disable=broad-except
-        error_handler(*sys.exc_info(), level=ERROR)
+        error_handler(*sys.exc_info(), level=logging.ERROR)
 
 
 @contextmanager
@@ -22,8 +20,8 @@ def unrecoverable(error_handler):
     try:
         yield
     except Exception as exc:  # pylint: disable=broad-except
-        error_handler(*sys.exc_info(), level=ERROR)
-        raise  # raise unrecoverableerror from x ?
+        error_handler(*sys.exc_info(), level=logging.ERROR)
+        raise  # raise unrecoverableerror from exc ?
 
 
 class Lifecycle:
@@ -55,12 +53,14 @@ class Lifecycle:
 
     @property
     def should_loop(self):
-        # TODO XXX started/stopped?
-        return not any((self.defunct, self.killed))
+        return self.alive and not any((self.defunct, self.killed))
 
     @property
     def status(self):
-        """One character status for this node. """
+        """
+        One character status for this node.
+
+        """
         if self._defunct:
             return '!'
         if not self.started:
@@ -97,9 +97,6 @@ class Lifecycle:
 
         self._stopped = True
 
-        if self._stopped:  # Stopping twice has no effect
-            return
-
     def kill(self):
         if not self.started:
             raise RuntimeError('Cannot kill an unstarted context.')
@@ -134,3 +131,12 @@ class BaseContext(Lifecycle, Wrapper):
         Lifecycle.__init__(self)
         Wrapper.__init__(self, wrapped)
         self.parent = parent
+
+    @property
+    def xstatus(self):
+        """
+        UNIX-like exit status, only coherent if the context has stopped.
+        """
+        if self._defunct:
+            return 70
+        return 0

bonobo/execution/contexts/graph.py

@@ -1,15 +1,25 @@
+import logging
 from functools import partial
+from queue import Empty
 from time import sleep
 
 from bonobo.config import create_container
 from bonobo.constants import BEGIN, END
+from bonobo.errors import InactiveReadableError
 from bonobo.execution import events
+from bonobo.execution.contexts.base import BaseContext
 from bonobo.execution.contexts.node import NodeExecutionContext
 from bonobo.execution.contexts.plugin import PluginExecutionContext
 from whistle import EventDispatcher
 
+logger = logging.getLogger(__name__)
 
-class GraphExecutionContext:
+
+class GraphExecutionContext(BaseContext):
+    """
+    Stores the actual state of a graph execution, and manages its lifecycle.
+
+    """
     NodeExecutionContextType = NodeExecutionContext
     PluginExecutionContextType = PluginExecutionContext
 
@@ -17,17 +27,24 @@ class GraphExecutionContext:
 
     @property
     def started(self):
+        if not len(self.nodes):
+            return super(GraphExecutionContext, self).started
         return any(node.started for node in self.nodes)
 
     @property
     def stopped(self):
+        if not len(self.nodes):
+            return super(GraphExecutionContext, self).stopped
         return all(node.started and node.stopped for node in self.nodes)
 
     @property
     def alive(self):
+        if not len(self.nodes):
+            return super(GraphExecutionContext, self).alive
         return any(node.alive for node in self.nodes)
 
-    def __init__(self, graph, plugins=None, services=None, dispatcher=None):
+    def __init__(self, graph, *, plugins=None, services=None, dispatcher=None):
+        super(GraphExecutionContext, self).__init__(graph)
         self.dispatcher = dispatcher or EventDispatcher()
         self.graph = graph
         self.nodes = [self.create_node_execution_context_for(node) for node in self.graph]
@@ -74,6 +91,8 @@ class GraphExecutionContext:
         self.dispatcher.dispatch(name, events.ExecutionEvent(self))
 
     def start(self, starter=None):
+        super(GraphExecutionContext, self).start()
+
         self.register_plugins()
         self.dispatch(events.START)
         self.tick(pause=False)
@@ -89,13 +108,21 @@ class GraphExecutionContext:
         if pause:
             sleep(self.TICK_PERIOD)
 
-    def kill(self):
-        self.dispatch(events.KILL)
-        for node_context in self.nodes:
-            node_context.kill()
-        self.tick()
+    def loop(self):
+        nodes = set(node for node in self.nodes if node.should_loop)
+        while self.should_loop and len(nodes):
+            self.tick(pause=False)
+            for node in list(nodes):
+                try:
+                    node.step()
+                except Empty:
+                    continue
+                except InactiveReadableError:
+                    nodes.discard(node)
 
     def stop(self, stopper=None):
+        super(GraphExecutionContext, self).stop()
+
         self.dispatch(events.STOP)
         for node_context in self.nodes:
             if stopper is None:
@@ -106,6 +133,14 @@ class GraphExecutionContext:
         self.dispatch(events.STOPPED)
         self.unregister_plugins()
 
+    def kill(self):
+        super(GraphExecutionContext, self).kill()
+
+        self.dispatch(events.KILL)
+        for node_context in self.nodes:
+            node_context.kill()
+        self.tick()
+
     def register_plugins(self):
         for plugin_context in self.plugins:
             plugin_context.register()
@@ -113,3 +148,11 @@ class GraphExecutionContext:
     def unregister_plugins(self):
         for plugin_context in self.plugins:
             plugin_context.unregister()
+
+    @property
+    def xstatus(self):
+        """
+        UNIX-like exit status, only coherent if the context has stopped.
+
+        """
+        return max(node.xstatus for node in self.nodes) if len(self.nodes) else 0

bonobo/execution/contexts/node.py

@@ -21,6 +21,16 @@ UnboundArguments = namedtuple('UnboundArguments', ['args', 'kwargs'])
 
 
 class NodeExecutionContext(BaseContext, WithStatistics):
+    """
+    Stores the actual context of a node, within a given graph execution, accessed as `self.parent`.
+
+    A special case exist, mostly for testing purpose, where there is no parent context.
+
+    Can be used as a context manager, also very convenient for testing nodes that requires some external context (like
+    a service implementation, or a value holder).
+
+    """
+
     def __init__(self, wrapped, *, parent=None, services=None, _input=None, _outputs=None):
         """
         Node execution context has the responsibility fo storing the state of a transformation during its execution.
@@ -77,10 +87,21 @@ class NodeExecutionContext(BaseContext, WithStatistics):
             initial = self._get_initial_context()
             self._stack = ContextCurrifier(self.wrapped, *initial.args, **initial.kwargs)
             if isconfigurabletype(self.wrapped):
+                try:
+                    self.wrapped = self.wrapped(_final=True)
+                except Exception as exc:
                     # Not normal to have a partially configured object here, so let's warn the user instead of having get into
                     # the hard trouble of understanding that by himself.
                     raise TypeError(
-                    'Configurables should be instanciated before execution starts.\nGot {!r}.\n'.format(self.wrapped)
+                        'Configurables should be instanciated before execution starts.\nGot {!r}.\n'.format(
+                            self.wrapped
+                        )
+                    ) from exc
+                else:
+                    raise TypeError(
+                        'Configurables should be instanciated before execution starts.\nGot {!r}.\n'.format(
+                            self.wrapped
+                        )
                     )
             self._stack.setup(self)
         except Exception:
@@ -102,9 +123,16 @@ class NodeExecutionContext(BaseContext, WithStatistics):
                 self.step()
             except InactiveReadableError:
                 break
+
+        logger.debug('Node loop ends for {!r}.'.format(self))
+
+    def step(self):
+        try:
+            self._step()
+        except InactiveReadableError:
+            raise
         except Empty:
             sleep(TICK_PERIOD)  # XXX: How do we determine this constant?
-                continue
         except (
                 NotImplementedError,
                 UnrecoverableError,
@@ -115,9 +143,7 @@ class NodeExecutionContext(BaseContext, WithStatistics):
         except BaseException:
             self.fatal(sys.exc_info())  # exit loop
 
-        logger.debug('Node loop ends for {!r}.'.format(self))
-
-    def step(self):
+    def _step(self):
         """
         A single step in the loop.
 
@@ -163,7 +189,7 @@ class NodeExecutionContext(BaseContext, WithStatistics):
         if self._stack:
             try:
                 self._stack.teardown()
-            except:
+            except Exception as exc:
                 self.fatal(sys.exc_info())
 
         super().stop()
@@ -231,12 +257,11 @@ class NodeExecutionContext(BaseContext, WithStatistics):
         :param mixed value: message
         """
         for message in messages:
-            if isinstance(message, Token):
+            if not isinstance(message, Token):
+                message = ensure_tuple(message, cls=self._input_type, length=self._input_length)
+                if self._input_length is None:
+                    self._input_length = len(message)
             self.input.put(message)
-            elif self._input_type:
-                self.input.put(ensure_tuple(message, cls=self._input_type))
-            else:
-                self.input.put(ensure_tuple(message))
 
     def write_sync(self, *messages):
         self.write(BEGIN, *messages, END)
@@ -264,17 +289,22 @@ class NodeExecutionContext(BaseContext, WithStatistics):
         If Queue raises (like Timeout or Empty), stat won't be changed.
 
         """
-        input_bag = self.input.get()
+        input_bag = self.input.get(timeout=0)
 
         # Store or check input type
         if self._input_type is None:
             self._input_type = type(input_bag)
-        elif type(input_bag) is not self._input_type:
+        elif type(input_bag) != self._input_type:
+            try:
+                if self._input_type == tuple:
+                    input_bag = self._input_type(input_bag)
+                else:
+                    input_bag = self._input_type(*input_bag)
+            except Exception as exc:
                 raise UnrecoverableTypeError(
-                'Input type changed between calls to {!r}.\nGot {!r} which is not of type {!r}.'.format(
-                    self.wrapped, input_bag, self._input_type
-                )
-            )
+                    'Input type changed to incompatible type between calls to {!r}.\nGot {!r} which is not of type {!r}.'.
+                    format(self.wrapped, input_bag, self._input_type)
+                ) from exc
 
         # Store or check input length, which is a soft fallback in case we're just using tuples
         if self._input_length is None:

bonobo/execution/strategies/executor.py

@@ -29,7 +29,7 @@ class ExecutorStrategy(Strategy):
         with self.create_executor() as executor:
             try:
                 context.start(self.get_starter(executor, futures))
-            except:
+            except Exception:
                 logger.critical('Exception caught while starting execution context.', exc_info=sys.exc_info())
 
             while context.alive:
@@ -53,14 +53,14 @@ class ExecutorStrategy(Strategy):
                 try:
                     with node:
                         node.loop()
-                except:
+                except Exception:
                     logging.getLogger(__name__).critical(
                         'Critical error in threadpool node starter.', exc_info=sys.exc_info()
                     )
 
             try:
                 futures.append(executor.submit(_runner))
-            except:
+            except Exception:
                 logging.getLogger(__name__).critical('futures.append', exc_info=sys.exc_info())
 
         return starter

bonobo/nodes/aggregation.py

@@ -0,0 +1,16 @@
+from bonobo.config import Configurable, Method, Option, ContextProcessor, use_raw_input
+from bonobo.util import ValueHolder
+
+
+class Reduce(Configurable):
+    function = Method()
+    initializer = Option(required=False)
+
+    @ContextProcessor
+    def buffer(self, context):
+        values = yield ValueHolder(self.initializer() if callable(self.initializer) else self.initializer)
+        context.send(values.get())
+
+    @use_raw_input
+    def __call__(self, values, bag):
+        values.set(self.function(values.get(), bag))

bonobo/nodes/io/csv.py

@@ -1,12 +1,11 @@
 import csv
 
-from bonobo.config import Option, use_raw_input, use_context
+from bonobo.config import Option, use_context
 from bonobo.config.options import Method, RenamedOption
 from bonobo.constants import NOT_MODIFIED
 from bonobo.nodes.io.base import FileHandler
 from bonobo.nodes.io.file import FileReader, FileWriter
 from bonobo.util import ensure_tuple
-from bonobo.util.bags import BagType
 
 
 class CsvHandler(FileHandler):

bonobo/plugins/sentry.py

@@ -1,5 +1,4 @@
 from bonobo.plugins import Plugin
-from raven import Client
 
 
 class SentryPlugin(Plugin):

bonobo/registry.py

@@ -89,6 +89,7 @@ class Registry:
 
 default_registry = Registry()
 
+
 def create_reader(name, *args, format=None, registry=default_registry, **kwargs):
     """
     Create a reader instance, guessing its factory using filename (and eventually format).
@@ -103,6 +104,7 @@ def create_reader(name, *args, format=None, registry=default_registry, **kwargs)
     """
     return registry.get_reader_factory_for(name, format=format)(name, *args, **kwargs)
 
+
 def create_writer(name, *args, format=None, registry=default_registry, **kwargs):
     """
     Create a writer instance, guessing its factory using filename (and eventually format).

bonobo/settings.py

@@ -52,7 +52,7 @@ class Setting:
     def set(self, value):
         value = self.formatter(value) if self.formatter else value
         if self.validator and not self.validator(value):
-            raise ValidationError('Invalid value {!r} for setting {}.'.format(value, self.name))
+            raise ValidationError(self, 'Invalid value {!r} for setting {!r}.'.format(value, self.name))
         self.value = value
 
     def set_if_true(self, value):

bonobo/util/collections.py

@@ -7,7 +7,23 @@ class sortedlist(list):
         bisect.insort(self, x)
 
 
-def ensure_tuple(tuple_or_mixed, *, cls=tuple):
+def _with_length_check(f):
+    @functools.wraps(f)
+    def _wrapped(*args, length=None, **kwargs):
+        nonlocal f
+        result = f(*args, **kwargs)
+        if length is not None:
+            if length != len(result):
+                raise TypeError(
+                    'Length check failed, expected {} fields but got {}: {!r}.'.format(length, len(result), result)
+                )
+        return result
+
+    return _wrapped
+
+
+@_with_length_check
+def ensure_tuple(tuple_or_mixed, *, cls=None):
     """
     If it's not a tuple, let's make a tuple of one item.
     Otherwise, not changed.
@@ -16,6 +32,8 @@ def ensure_tuple(tuple_or_mixed, *, cls=tuple):
     :return: tuple
 
     """
+    if cls is None:
+        cls = tuple
 
     if isinstance(tuple_or_mixed, cls):
         return tuple_or_mixed

bonobo/util/environ.py

@@ -57,7 +57,6 @@ def get_argument_parser(parser=None):
     :return:
     """
     if parser is None:
-        import argparse
         parser = argparse.ArgumentParser()
 
     # Store globally to be able to warn the user about the fact he's probably wrong not to pass a parser to

requirements-dev.txt

@@ -1,6 +1,6 @@
 -e .[dev]
 -r requirements.txt
-alabaster==0.7.10
+alabaster==0.7.11
 arrow==0.12.1
 atomicwrites==1.1.5
 attrs==18.1.0
@@ -13,7 +13,7 @@ cookiecutter==1.5.1
 coverage==4.5.1
 docutils==0.14
 future==0.16.0
-idna==2.6
+idna==2.7
 imagesize==1.0.0
 jinja2-time==0.2.0
 jinja2==2.10
@@ -22,20 +22,20 @@ more-itertools==4.2.0
 packaging==17.1
 pluggy==0.6.0
 poyo==0.4.1
-py==1.5.3
+py==1.5.4
 pygments==2.2.0
 pyparsing==2.2.0
 pytest-cov==2.5.1
-pytest-timeout==1.2.1
-pytest==3.6.1
+pytest-timeout==1.3.0
+pytest==3.6.3
 python-dateutil==2.7.3
-pytz==2018.4
-requests==2.18.4
+pytz==2018.5
+requests==2.19.1
 six==1.11.0
 snowballstemmer==1.2.1
 sphinx-sitemap==0.2
-sphinx==1.7.5
+sphinx==1.7.6
 sphinxcontrib-websupport==1.1.0
-urllib3==1.22
+urllib3==1.23
 whichcraft==0.4.1
 yapf==0.22.0

requirements-docker.txt

@@ -7,24 +7,23 @@ chardet==3.0.4
 colorama==0.3.9
 docker-pycreds==0.3.0
 docker==2.7.0
-fs==2.0.23
-graphviz==0.8.3
-idna==2.6
+fs==2.0.25
+graphviz==0.8.4
+idna==2.7
 jinja2==2.10
 markupsafe==1.0
 mondrian==0.7.0
 packaging==17.1
-pbr==4.0.4
+pbr==4.1.1
 psutil==5.4.6
 pyparsing==2.2.0
 python-slugify==1.2.5
-pytz==2018.4
-requests==2.18.4
+pytz==2018.5
+requests==2.19.1
 semantic-version==2.6.0
 six==1.11.0
-stevedore==1.28.0
-typing==3.6.4
+stevedore==1.29.0
 unidecode==1.0.22
-urllib3==1.22
+urllib3==1.23
 websocket-client==0.48.0
 whistle==1.0.1

requirements-jupyter.txt

@@ -10,7 +10,7 @@ ipykernel==4.8.2
 ipython-genutils==0.2.0
 ipython==6.4.0
 ipywidgets==6.0.1
-jedi==0.12.0
+jedi==0.12.1
 jinja2==2.10
 jsonschema==2.6.0
 jupyter-client==5.2.3
@@ -21,23 +21,24 @@ markupsafe==1.0
 mistune==0.8.3
 nbconvert==5.3.1
 nbformat==4.4.0
-notebook==5.5.0
+notebook==5.6.0
 pandocfilters==1.4.2
-parso==0.2.1
+parso==0.3.1
 pexpect==4.6.0
 pickleshare==0.7.4
+prometheus-client==0.3.0
 prompt-toolkit==1.0.15
-ptyprocess==0.5.2
+ptyprocess==0.6.0
 pygments==2.2.0
 python-dateutil==2.7.3
-pyzmq==17.0.0
+pyzmq==17.1.0
 qtconsole==4.3.1
 send2trash==1.5.0
 simplegeneric==0.8.1
 six==1.11.0
 terminado==0.8.1
 testpath==0.3.1
-tornado==5.0.2
+tornado==5.1
 traitlets==4.3.2
 wcwidth==0.1.7
 webencodings==0.5.1

requirements-sqlalchemy.txt

@@ -5,23 +5,22 @@ bonobo-sqlalchemy==0.6.0
 certifi==2018.4.16
 chardet==3.0.4
 colorama==0.3.9
-fs==2.0.23
-graphviz==0.8.3
-idna==2.6
+fs==2.0.25
+graphviz==0.8.4
+idna==2.7
 jinja2==2.10
 markupsafe==1.0
 mondrian==0.7.0
 packaging==17.1
-pbr==4.0.4
+pbr==4.1.1
 psutil==5.4.6
 pyparsing==2.2.0
 python-slugify==1.2.5
-pytz==2018.4
-requests==2.18.4
+pytz==2018.5
+requests==2.19.1
 six==1.11.0
-sqlalchemy==1.2.8
-stevedore==1.28.0
-typing==3.6.4
+sqlalchemy==1.2.10
+stevedore==1.29.0
 unidecode==1.0.22
-urllib3==1.22
+urllib3==1.23
 whistle==1.0.1

requirements.txt

@@ -3,22 +3,21 @@ appdirs==1.4.3
 certifi==2018.4.16
 chardet==3.0.4
 colorama==0.3.9
-fs==2.0.23
-graphviz==0.8.3
-idna==2.6
+fs==2.0.25
+graphviz==0.8.4
+idna==2.7
 jinja2==2.10
 markupsafe==1.0
 mondrian==0.7.0
 packaging==17.1
-pbr==4.0.4
+pbr==4.1.1
 psutil==5.4.6
 pyparsing==2.2.0
 python-slugify==1.2.5
-pytz==2018.4
-requests==2.18.4
+pytz==2018.5
+requests==2.19.1
 six==1.11.0
-stevedore==1.28.0
-typing==3.6.4
+stevedore==1.29.0
 unidecode==1.0.22
-urllib3==1.22
+urllib3==1.23
 whistle==1.0.1

setup.py

@@ -1,4 +1,4 @@
-# Generated by Medikit 0.6.3 on 2018-06-11.
+# Generated by Medikit 0.6.3 on 2018-07-22.
 # All changes will be overriden.
 # Edit Projectfile and run “make update” (or “medikit update”) to regenerate.
 

tests/execution/contexts/test_execution_contexts_graph.py

@@ -0,0 +1,59 @@
+from bonobo import Graph
+from bonobo.constants import EMPTY, BEGIN, END
+from bonobo.execution.contexts import GraphExecutionContext
+
+
+def raise_an_error(*args, **kwargs):
+    raise Exception('Careful, man, there\'s a beverage here!')
+
+
+def raise_an_unrecoverrable_error(*args, **kwargs):
+    raise Exception('You are entering a world of pain!')
+
+
+def test_lifecycle_of_empty_graph():
+    graph = Graph()
+    with GraphExecutionContext(graph) as context:
+        assert context.started
+        assert context.alive
+        assert not context.stopped
+    assert context.started
+    assert not context.alive
+    assert context.stopped
+    assert not context.xstatus
+
+
+def test_lifecycle_of_nonempty_graph():
+    graph = Graph([1, 2, 3], print)
+    with GraphExecutionContext(graph) as context:
+        assert context.started
+        assert context.alive
+        assert not context.stopped
+    assert context.started
+    assert not context.alive
+    assert context.stopped
+    assert not context.xstatus
+
+
+def test_lifecycle_of_graph_with_recoverable_error():
+    graph = Graph([1, 2, 3], raise_an_error, print)
+    with GraphExecutionContext(graph) as context:
+        assert context.started
+        assert context.alive
+        assert not context.stopped
+    assert context.started
+    assert not context.alive
+    assert context.stopped
+    assert not context.xstatus
+
+
+def test_lifecycle_of_graph_with_unrecoverable_error():
+    graph = Graph([1, 2, 3], raise_an_unrecoverrable_error, print)
+    with GraphExecutionContext(graph) as context:
+        assert context.started and context.alive and not context.stopped
+        context.write(BEGIN, EMPTY, END)
+        context.loop()
+    assert context.started
+    assert not context.alive
+    assert context.stopped
+    assert not context.xstatus