Allow n_jobs to receive negative values.

tsfresh/convenience/relevant_extraction.py

@@ -98,7 +98,11 @@ def extract_relevant_features(
         smaller chunksize.
     :type chunksize: None or int
 
-    :param n_jobs: The number of processes to use for parallelization. If zero, no parallelization is used.
+    :param n_jobs: The number of processes to use for parallelization.
+                If zero, no parallelization is used.
+                ``-1`` means using all processors. See scikit-learns'
+                `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                for more details.
     :type n_jobs: int
 
     :param distributor: Advanced parameter: set this to a class name that you want to use as a

tsfresh/feature_extraction/extraction.py

@@ -21,6 +21,7 @@
     DistributorBaseClass,
     MapDistributor,
     MultiprocessingDistributor,
+    effective_n_jobs,
 )
 from tsfresh.utilities.string_manipulation import convert_to_output_format
 
@@ -100,7 +101,11 @@ def extract_features(
     :param column_value: The name for the column keeping the value itself. Please see :ref:`data-formats-label`.
     :type column_value: str
 
-    :param n_jobs: The number of processes to use for parallelization. If zero, no parallelization is used.
+    :param n_jobs: The number of processes to use for parallelization.
+                If zero, no parallelization is used.
+                ``-1`` means using all processors. See scikit-learns'
+                `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                for more details.
     :type n_jobs: int
 
     :param chunksize: The size of one chunk that is submitted to the worker
@@ -240,7 +245,11 @@ def _do_extraction(
     :param chunk_size: The size of one chunk for the parallelization
     :type chunk_size: None or int
 
-    :param n_jobs: The number of processes to use for parallelization. If zero, no parallelization is used.
+    :param n_jobs: The number of processes to use for parallelization.
+                If zero, no parallelization is used.
+                ``-1`` means using all processors. See scikit-learns'
+                `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                for more details.
     :type n_jobs: int
 
     :param disable_progressbar: Do not show a progressbar while doing the calculation.
@@ -258,6 +267,7 @@ def _do_extraction(
     """
 
     data = to_tsdata(df, column_id, column_kind, column_value, column_sort)
+    n_jobs = effective_n_jobs(n_jobs)
 
     if distributor is None:
         if isinstance(data, Iterable):

tsfresh/feature_selection/relevance.py

@@ -25,7 +25,10 @@
     target_real_feature_binary_test,
     target_real_feature_real_test,
 )
-from tsfresh.utilities.distribution import initialize_warnings_in_workers
+from tsfresh.utilities.distribution import (
+    initialize_warnings_in_workers,
+    effective_n_jobs
+)
 
 
 def calculate_relevance_table(
@@ -128,7 +131,11 @@ def calculate_relevance_table(
                                    independent (e.g. mean and median)
     :type hypotheses_independent: bool
 
-    :param n_jobs: Number of processes to use during the p-value calculation
+    :param n_jobs: Number of processes to use during the p-value calculation.
+                If zero, no parallelization is used.
+                ``-1`` means using all processors. See scikit-learns'
+                `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                for more details.
     :type n_jobs: int
 
     :param show_warnings: Show warnings during the p-value calculation (needed for debugging of calculators).
@@ -192,6 +199,8 @@ def calculate_relevance_table(
         else:
             warnings.simplefilter("default")
 
+        n_jobs = effective_n_jobs(n_jobs)
+
         if n_jobs == 0 or n_jobs == 1:
             map_function = map
         else:

tsfresh/feature_selection/selection.py

@@ -108,7 +108,11 @@ def select_features(
                                    independent (e.g. mean and median)
     :type hypotheses_independent: bool
 
-    :param n_jobs: Number of processes to use during the p-value calculation
+    :param n_jobs: Number of processes to use during the p-value calculation.
+                If zero, no parallelization is used.
+                ``-1`` means using all processors. See scikit-learns'
+                `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                for more details.
     :type n_jobs: int
 
     :param show_warnings: Show warnings during the p-value calculation (needed for debugging of calculators).

tsfresh/transformers/feature_augmenter.py

@@ -101,7 +101,11 @@ def __init__(
         :param column_value: The column with the values. See :mod:`~tsfresh.feature_extraction.extraction`.
         :type column_value: basestring
 
-        :param n_jobs: The number of processes to use for parallelization. If zero, no parallelization is used.
+        :param n_jobs: The number of processes to use for parallelization.
+                    If zero, no parallelization is used.
+                    ``-1`` means using all processors. See scikit-learns'
+                    `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                    for more details.
         :type n_jobs: int
 
         :param chunksize: The size of one chunk that is submitted to the worker

tsfresh/transformers/feature_selector.py

@@ -99,7 +99,11 @@ def __init__(
                                        independent (e.g. mean and median)
         :type hypotheses_independent: bool
 
-        :param n_jobs: Number of processes to use during the p-value calculation
+        :param n_jobs: Number of processes to use during the p-value calculation.
+                    If zero, no parallelization is used.
+                    ``-1`` means using all processors. See scikit-learns'
+                    `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                    for more details.
         :type n_jobs: int
 
         :param chunksize: Size of the chunks submitted to the worker processes

tsfresh/transformers/relevant_feature_augmenter.py

@@ -154,7 +154,11 @@ def __init__(
             smaller chunksize.
         :type chunksize: None or int
 
-        :param n_jobs: The number of processes to use for parallelization. If zero, no parallelization is used.
+        :param n_jobs: The number of processes to use for parallelization.
+                    If zero, no parallelization is used.
+                    ``-1`` means using all processors. See scikit-learns'
+                    `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                    for more details.
         :type n_jobs: int
 
         :param show_warnings: Show warnings during the feature extraction (needed for debugging of calculators).

tsfresh/utilities/dataframe_functions.py

@@ -15,6 +15,7 @@
     DistributorBaseClass,
     MapDistributor,
     MultiprocessingDistributor,
+    effective_n_jobs,
 )
 
 
@@ -420,7 +421,11 @@ def roll_time_series(
          than or equal 0.
     :type min_timeshift: int
 
-    :param n_jobs: The number of processes to use for parallelization. If zero, no parallelization is used.
+    :param n_jobs: The number of processes to use for parallelization.
+                If zero, no parallelization is used.
+                ``-1`` means using all processors. See scikit-learns'
+                `Glossary <https://scikit-learn.org/stable/glossary.html#term-n-jobs>`__
+                for more details.
     :type n_jobs: int
 
     :param chunksize: How many shifts per job should be calculated.
@@ -541,6 +546,7 @@ def roll_time_series(
         range_of_shifts = range(1, prediction_steps + 1, rolling_amount)
 
     if distributor is None:
+        n_jobs = effective_n_jobs(n_jobs)
         if n_jobs == 0 or n_jobs == 1:
             distributor = MapDistributor(
                 disable_progressbar=disable_progressbar, progressbar_title="Rolling"

tsfresh/utilities/distribution.py

@@ -14,7 +14,7 @@
 from collections.abc import Generator, Iterable
 from functools import partial
 from itertools import islice, repeat, takewhile
-from multiprocessing import Pool
+from multiprocessing import Pool, cpu_count
 
 from tqdm import tqdm
 
@@ -61,6 +61,13 @@ def initialize_warnings_in_workers(show_warnings):
         warnings.simplefilter("default")
 
 
+def effective_n_jobs(n_jobs):
+    if n_jobs < 0:
+        n_jobs = max(cpu_count() + 1 + n_jobs, 1)
+
+    return n_jobs
+
+
 class DistributorBaseClass:
     """
     The distributor abstract base class.