Grid2op

---------------------
- [FIXED] forgot to print the name of the missing environment when error in creating it.
- [FIXED] an issue in `MultiFolder.sample_next_chronics` that did not returns the right index
- [FIXED] an issue that prevented the `EpisodeData` class to load back properly the action of the environment.
This might have side effect if you used the `obs.from_vect` or `act.from_vect` in non conventional ways.
- [ADDED] some documentation and example for the `MultiProcessEnv`
- [IMPROVED] check that the sub environments are suitable grid2op.Environment.Environment in multiprocess env.
- [FIXED] Minor documentation generation warnings and typos (Parameters, Backend, OpponentSpace, ActionSpace)

---------------------
- [BREAKING] `MultiEnv` has been renamed `SingleEnvMultiProcess`
- [BREAKING] `MultiEnv` has been abstracted to `BaseMultiProcessEnv` and the backwards compatible interface is now
`SingleProcessMultiEnv`
- [BREAKING] the `seeds` parameters of the `Runner.run` function has been renamed `env_seeds` and an `agent_seeds`
parameters is now available for fully reproducible experiments.
- [FIXED] a weird effect on `env.reset` that did not reset the state of the previous observation held
by the environment. This could have caused some issue in some corner cases.
- [FIXED] `BaseAction.__iadd__` fixed a bug with change actions `+=` operator reported in
`Issue 116 <https://github.com/Grid2Op/grid2op/issues/116>`_
- [FIXED] `obs.simulate` post-initialized reward behaves like the environment
- [FIXED] `LinesReconnectedReward` fixes reward inverted range
- [FIXED] the `get_all_unitary_topologies_change` now counts only once the "do nothing" action.
- [FIXED] `obs.simulate` could sometime returns "None" when the simulated action lead to a game over. This is no longer
a problem.
- [FIXED] `grid2op.make` will now raise an error if an invalid argument has been passed to it.
- [FIXED] some arguments were not passed correctly to `env.get_kwargs()` or `env.get_params_for_runner()`
- [ADDED] `Issue 110 <https://github.com/Grid2Op/grid2op/issues/110>`_ Adding an agent that is able to reconnect
disconnected powerlines that can be reconnected, see `grid2op.Agent.RecoPowerlineAgent`
- [ADDED] a clearer explanation between illegal and ambiguous action.
- [ADDED] `MultiEnvMultiProcess` as a new multi-process class to run different environments in multiples prallel
processes.
- [ADDED] more control on the environment when using the `grid2op.make` function.
- [ADDED] creation of the MultiMixEnv that allows to have, through a unified interface the possibility to interact
alternatively with one environment or the other. This is especially useful when considering an agent that should
interact in multiple environments.
- [ADDED] possibility to use `simulate` on the current observation.
- [ADDED] the overload of "__getattr__" for environment running in parallel
- [ADDED] capability to change the powerlines on which the opponent attack at the environment initialization
- [UPDATED] `Backend.PandaPowerBackend.apply_action` vectorized backend apply action method for speed.
- [UPDATED] `Issue 111 <https://github.com/Grid2Op/grid2op/issues/111>`_ Converter is better documented to be
more broadly usable.
- [UPDATED] `MultiEnv` has been updated for new use case: Providing different environments configurations on the same
grid and an arbitrary number of processes for each of these.
- [UPDATED] Behaviour of "change_bus" and "set_bus": it is no more possible to affect the bus of a powerline
disconnected.
- [UPDATED] More control about the looping strategy of the `ChronicsHandler` that has been refactored, and can now be
more easily cached (no need to do an expensive reading of the data at each call to `env.reset`)

---------------------
- [FIXED] `Issue 114 <https://github.com/Grid2Op/grid2op/issues/114>`_ the issue concerning the
bug for the maintenance.

---------------------
- [FIXED] `Issue 69 <https://github.com/Grid2Op/grid2op/issues/69>`_ MultEnvironment is now working with windows
based OS.
- [ADDED] `Issue 108 <https://github.com/Grid2Op/grid2op/issues/108>`_ Seed is now part of the public agent API.
The notebook has been updated accordingly.
- [ADDED] Some function to disable the `obs.simulate` if wanted. This can lead to around 10~15% performance speed up
in case `obs.simulate` is not used. See `env.deactivate_forecast` and `env.reactivate_forecast`
(related to `Issued 98 <https://github.com/Grid2Op/grid2op/issues/98>`_)
- [UPDATED] the first introductory notebook.
- [UPDATED] possibility to reconnect / disconnect powerline giving its name when using `reconnect_powerline` and
`disconnect_powerline` methods of the action space.
- [UPDATED] `Issue 105 <https://github.com/Grid2Op/grid2op/issues/105>`_ problem solved for notebook 4.
based OS.
- [UPDATED] overall speed enhancement mostly in the `VoltageControler`, with the adding of the previous capability,
some updates in the `BackendAction`
`Issued 98 <https://github.com/Grid2Op/grid2op/issues/98>`_
- [UPDATED] Added `PlotMatplot` constructor arguments to control display of names and IDs of the grid elements
(gen, load, lines). As suggested in `Issue 106 <https://github.com/Grid2Op/grid2op/issues/106>`_

---------------------
- [FIXED] `GridObject` loading from file does initialize single values (`bool`, `int`, `float`)
correctly instead of creating a `np.array` of size one.
- [FIXED] `IdToAct` loading actions from file .npy
- [FIXED] a problem on the grid name import on some version of pandas
- [ADDED] a function that returns the types of the action see `action.get_types()`
- [ADDED] a class to "cache" the data in memory instead of reading it over an over again from disk (see
`grid2op.chronics.MultifolderWithCache` (related to
`Issued 98 <https://github.com/Grid2Op/grid2op/issues/98>`_) )
- [ADDED] improve the documentation of the observation class.
- [UPDATED] Reward `LinesReconnectedReward` to take into account maintenances downtimes
- [UPDATED] Adds an option to disable plotting load and generators names when using `PlotMatplot`

---------------------
- [FIXED] a bug preventing to save gif with episode replay when there has been a game over before starting time step
- [FIXED] the issue of the random seed used in the environment for the runner.