Childproject

Added

- add the derived annotation pipeline
- audio conversion has now a 'standard' conversion pipeline that will convert files to mono channel 16kHz pcm_s16le (no need to convert channels and then sampling rate and no need to know the options)
- add the simple_CTC metric to the list of available metrics

Changed

- the output of the CLI in the terminal is now handled by the logger module and not by print statements
- validating a dataset now results in warnings for broken symlinks and no errors anymore (425)
- validation with recordings existing but for which mediainfo can't read the sample rate no longer fail but outputs a warning.
- recording_device_type in recordings.csv now accepts izyrec
- rework of the list of available metrics functions: now has a absolute value and per_hour decorator, and a peak_ option

Fixed

- zooniverse chunks upload was failing if the dataset column was missing in the csv
- eaf_builder with speaker_id NA no longer fails (438)
- name of files in rttm that contain only digits in their name now work correctly with the filter ($457)
- periodic sampler having extra NA rows when using explode

Added

- validation of the annotation index checks for annotation period outside of audio duration

Changed

- ignore_discarded is now the default behaviour of childproject projects, robustness was added to the discard column
- annotation index validation and annotation importation now checks for the range_offset being greater than duration and oiutputs an error when it is the case
- projects now check for unicity of the experiment column in children and recording csv, read() fails when not unique
- zooniverse uploads : uploads extra metadata to zooniverse (name of the audio clip, dataset it belongs to)
- zooniverse uploads : the subject_set id is stored in the chunk csv file, as the subject_set name (display name) is susceptible to change
- zooniverse uploads : the upload now handles SIGINT and SIGTERM signals to save progression of the upload to the csv before exiting (useful when a job needs to be interrupted
- allow once again get_within_time_range to take str arguments as times
- add arguments to choose the format of compute_ages project method

Fixed

- discard column in recordings.csv and children.csv now works properly
- metrics pipeline now checks the converted name for unicity even if a specific name was given
- rename set also renames the merged_from column
- rename set accepts subsets location without failing

Added

- Windows automated tests (some functions were edited to be windows compatible). TO REMEMBER: type int in windows default to int32 instead of int64, can lead to big int turned into negative values
- Check calculated ages during corpus validation

Changed

- pandas version restricted to avoid errors of future releases , (1.1.0 (assert_frame_equal check_less_precise) to 1.5.0 (last checked version))
- no usage of sox command anymore, remove sox dependency
- merging annotations now sets the format to 'NA' instead of a blank value.

Fixed

- replace exit() with raise ValueError() to comply with Exception propagation (metrics pipeline)
- fixed ignore-errors in zooniverse upload_chunks
- fixed calculate_shift to correctly reshape to single channel