PYME.recipes.recipe module

class PYME.recipes.recipe.Recipe(*args, **kwargs)

Bases: HasTraits

add_module(module)

add_modules_and_execute(modules, rollback_on_failure=True)

Adds modules to the recipe and then execute the recipe. Added to make UI interaction in PYMEVis a bit nicer when modules added from the menu fail, this function gives the option (enabled by default) of rolling back the additions should execute fail.

Parameters

moduleslist

a list of modules to add

rollback_on_failurebool

rollback and remove modules (and their outputs) if .execute() fails

clear()

dependancyGraph()

downstream_outputs(keys)

execute(progress_callback=None, **kwargs)

Execute the recipe. Recipe execution is lazy / incremental in that modules will only be executed if their outputs are no longer in the namespace or their inputs have changed.

Parameters

progress_callbacka function, progress_callback(recipe, module)

This function is called after the successful execution of each module in the recipe. To abort the recipe, an exception may be raised in progress_callback.

kwargsany values to set in the recipe namespace prior to executing.

Returns

outputthe value of the ‘output’ variable in the namespace if present, otherwise none. output is only used in

dh5view and ignored in recipe batch processing. It might well disappear completely in the future.

execute_on_invalidation = <PYME.misc.mock_traits.CStr object>

property file_inputs

classmethod fromJSON(data)

classmethod fromMD(md)

classmethod fromYAML(data)

gather_outputs(context={})

Find all OutputModule instances and call their generate methods with the recipe context

Parameters

contextdict

A context dictionary used to substitute and create variable names.

get_cleaned_module_list()

property inputs

invalidate_data()

layout()

Added as a visualisation aid for the web-based recipe editor. Very much a work in progress and not guaranteed to remain in it’s current form.

TODO - does this logic belong here???? TODO - rename?? (potentially make it webui specific)??? TODO - potential issues on Py3 with how jigna treats namedtuple?

loadInput(filename, key='input', metadata_defaults={})

Load input data from a file and inject into namespace

property module_outputs

modules = <PYME.misc.mock_traits.List object>

new_output_name(stub)

property outputs

property pipeline_view

prune_dependencies_from_namespace(keys_to_prune, keep_passed_keys=False)

resolveDependencies()

reverseDependancyGraph()

save(context={})

Find all OutputModule instances and call their save methods with the recipe context

Parameters

contextdict

A context dictionary used to substitute and create variable names.

save_yaml(uri)

Save the recipe text to .yaml using the cluster-aware unified IO library

WARNING: Experimental - this was added as a quick hack to get web-based recipe editing working, and WILL LIKELY BE REMOVED without deprecation once that moves to using a recipe-manager. Whilst arguably the least contentious of the web-editor additions, it is unclear whether the saving logic (past dumping to YAML in the toYAML() method) should reside within the recipe itself. As the recipe class is proxied into the browser, there are also potential security implications here, particularly as this accepts both filenames and clusterURIs. As a consequence, saving should probably be factored out into something which can operate in an appropriate sandbox (the other option is to sandbox unifiedIO).

Parameters

uri: str

A filename or PYME-CLUSTER:// URI

toJSON()

toYAML()

to_html(show_progress=True, show_errors=True)

to_svg()

update_from_file(filename)

Update the contents of the recipe from a .yaml file

WARNING: This function will likely be REMOVED WITHOUT NOTICE. It is a quick hack to get the prototype web-based recipe editor working, but will be surplus to requirements once we have a proper recipe manager in the web based editor. It’s logically obtuse to consider something the same recipe once you’ve completely replaced it with a recipe that has been loaded from file. It is much more sensible to create a new recipe instance when loading a recipe from file, and this is the recommended approach.

Parameters

filename: str

filename or PYME-CLUSTER:// URI

update_from_yaml(data)

Update from a yaml formatted recipe description

Parameters

data: str

either yaml formatted text, or the path to a yaml file.

Returns

None

upstream_inputs(keys)

exception PYME.recipes.recipe.RecipeError

Bases: Exception

exception PYME.recipes.recipe.RecipeExecutionError(msg, recipe=None)

Bases: RecipeError

exception PYME.recipes.recipe.RecipeModuleNotFoundError

Bases: RecipeError