Dynamically dispatch function but do not print result

[albertotb]

Can Fire be used to dynamically dispatch a function as in the dict example but without printing the result? I would like to do something with the return value, for instance:

if __name__ == "__main__":
    # inception and exception are functions that return keras models, 
    # each having its own parameters 
    ret = Fire.Fire({'inception': inception,
                     'xception': xception})

    # do something with ret

If I try to run the previous code, for instance:

python ./train.py inception 100 100 10

I get the following error:

ERROR: The function received no value for the required argument: inputs
2020-03-03 16:30:13.774542: W tensorflow/python/util/util.cc:319] Sets are not currently considered sequences, but this may change in the future, so consider avoiding using them.
Usage: train.py inception 100 100 10 - <group|command|value> | --inputs=INPUTS <flags> [ARGS]...
  available groups:      inbound_nodes | input | input_names | input_shape |
                         inputs | layers | losses | metrics | metrics_names |
                         non_trainable_variables | non_trainable_weights |
                         outbound_nodes | output | output_names |
                         output_shape | outputs | state_updates | submodules |
                         trainable_variables | trainable_weights | updates |
                         variables | weights
  available commands:    add_loss | add_metric | add_update | add_variable |
                         add_weight | apply | build | call | compile |
                         compute_mask | compute_output_shape |
                         compute_output_signature | count_params | evaluate |
                         evaluate_generator | fit | fit_generator |
                         from_config | get_config | get_input_at |
                         get_input_mask_at | get_input_shape_at | get_layer |
                         get_losses_for | get_output_at | get_output_mask_at |
                         get_output_shape_at | get_updates_for | get_weights |
                         load_weights | name_scope | predict |
                         predict_generator | predict_on_batch | reset_metrics |
                         reset_states | save | save_weights | set_weights |
                         summary | test_on_batch | to_json | to_yaml |
                         train_on_batch | with_name_scope
  available values:      activity_regularizer | built | dtype | dynamic |
                         input_mask | input_spec | name | optimizer |
                         output_mask | run_eagerly | stateful |
                         supports_masking | trainable
  flags are accepted

For detailed information on this command, run:
  train.py inception 100 100 10 - -- --help`

[dbieber]

This is not an option today but is a good idea. Thanks for opening the issue.

[dbieber]

Take a look at the discussion in #188, which includes proposals for how to disable printing the output.

[spyoungtech]

One way you might be able to do something like this is by using contextlib.redirect_stdout to suppress the output. For example, you could redirect to /dev/null (or platform equivalent)

import os
import contextlib
with open(os.devnull, 'w') as devnull:
    with contextlib.redirect_stdout(devnull):
        ret = fire.Fire(something)
# do something with ret