Move conversion functions to their own modules and remove cyclic imports (#370)

* Move wav, edf, mat, and csv conversion functions to their own modules

* Move content to remove cyclic dependencies

* Move get_version for a PhysioNet database, to download.py

* Remove rdrecord's ability to read edf and wav files

* Update documentation

Author: cx1111

docs/convert.rst

@@ -0,0 +1,35 @@
+Format Conversions
+==================
+
+The wfdb.io.convert subpackage contains functions to read and write
+non-WFDB files commonly used for waveforms.
+
+CSV
+---
+
+.. automodule:: wfdb.io.convert.csv
+   :members: csv_to_wfdb
+
+EDF
+---
+
+.. automodule:: wfdb.io.convert.edf
+   :members:
+
+Matlab
+------
+
+.. automodule:: wfdb.io.convert.matlab
+   :members:
+
+TFF
+---
+
+.. automodule:: wfdb.io.convert.tff
+   :members:
+
+WAV
+---
+
+.. automodule:: wfdb.io.convert.wav
+   :members:

docs/index.rst

@@ -53,6 +53,7 @@ Other Content
 
    installation
    wfdb-specifications
+   convert
 
 
 Indices and tables

docs/io.rst

@@ -9,7 +9,7 @@ WFDB Records
 ---------------
 
 .. automodule:: wfdb.io
-    :members: rdrecord, rdheader, rdsamp, wrsamp, edf2mit, mit2edf, wav2mit, mit2wav, csv2mit
+    :members: rdrecord, rdheader, rdsamp, wrsamp
 
 .. autoclass:: wfdb.io.Record
     :members: wrsamp, adc, dac
@@ -32,4 +32,4 @@ Downloading
 -----------
 
 .. automodule:: wfdb.io
-    :members: get_dbs, get_record_list, dl_database, dl_files, set_db_index_url
+    :members: dl_files, dl_database, get_dbs, get_record_list, set_db_index_url

docs/processing.rst

@@ -17,7 +17,7 @@ Heart Rate
 ----------
 
 .. automodule:: wfdb.processing
-    :members: compute_hr
+    :members: compute_hr, calc_rr, calc_mean_hr, ann2rr, rr2ann
 
 
 Peaks
@@ -26,6 +26,11 @@ Peaks
 .. automodule:: wfdb.processing
     :members: find_peaks, find_local_peaks, correct_peaks
 
+Filters
+-------
+
+.. automodule:: wfdb.processing
+    :members: sigavg
 
 QRS Detectors
 -------------

docs/wfdb.rst

@@ -9,7 +9,7 @@ WFDB Records
 ---------------
 
 .. automodule:: wfdb
-    :members: rdrecord, rdheader, rdsamp, wrsamp, edf2mit, mit2edf, wav2mit, mit2wav, csv2mit
+    :members: rdrecord, rdheader, rdsamp, wrsamp
 
 .. autoclass:: wfdb.Record
     :members: wrsamp, adc, dac
@@ -32,7 +32,7 @@ Downloading
 -----------
 
 .. automodule:: wfdb
-    :members: get_dbs, get_record_list, dl_database, dl_files, set_db_index_url
+    :members: dl_files, dl_database, get_dbs, get_record_list, set_db_index_url
 
 
 Plotting

tests/io/test_convert.py

@@ -0,0 +1,110 @@
+import numpy as np
+
+from wfdb.io.record import rdrecord
+from wfdb.io.convert.edf import read_edf
+
+
+class TestConvert:
+    def test_edf_uniform(self):
+        """
+        EDF format conversion to MIT for uniform sample rates.
+
+        """
+        # Uniform sample rates
+        record_MIT = rdrecord("sample-data/n16").__dict__
+        record_EDF = read_edf("sample-data/n16.edf").__dict__
+
+        fields = list(record_MIT.keys())
+        # Original MIT format method of checksum is outdated, sometimes
+        # the same value though
+        fields.remove("checksum")
+        # Original MIT format units are less comprehensive since they
+        # default to mV if unknown.. therefore added more default labels
+        fields.remove("units")
+
+        test_results = []
+        for field in fields:
+            # Signal value will be slightly off due to C to Python type conversion
+            if field == "p_signal":
+                true_array = np.array(record_MIT[field]).flatten()
+                pred_array = np.array(record_EDF[field]).flatten()
+                # Prevent divide by zero warning
+                for i, v in enumerate(true_array):
+                    if v == 0:
+                        true_array[i] = 1
+                        pred_array[i] = 1
+                sig_diff = np.abs((pred_array - true_array) / true_array)
+                sig_diff[sig_diff == -np.inf] = 0
+                sig_diff[sig_diff == np.inf] = 0
+                sig_diff = np.nanmean(sig_diff, 0)
+                # 5% tolerance
+                if np.max(sig_diff) <= 5:
+                    test_results.append(True)
+                else:
+                    test_results.append(False)
+            elif field == "init_value":
+                signal_diff = [
+                    abs(record_MIT[field][i] - record_EDF[field][i])
+                    for i in range(len(record_MIT[field]))
+                ]
+                if abs(max(min(signal_diff), max(signal_diff), key=abs)) <= 2:
+                    test_results.append(True)
+                else:
+                    test_results.append(False)
+            else:
+                test_results.append(record_MIT[field] == record_MIT[field])
+
+        target_results = len(fields) * [True]
+        assert np.array_equal(test_results, target_results)
+
+    def test_edf_non_uniform(self):
+        """
+        EDF format conversion to MIT for non-uniform sample rates.
+
+        """
+        # Non-uniform sample rates
+        record_MIT = rdrecord("sample-data/wave_4").__dict__
+        record_EDF = read_edf("sample-data/wave_4.edf").__dict__
+
+        fields = list(record_MIT.keys())
+        # Original MIT format method of checksum is outdated, sometimes
+        # the same value though
+        fields.remove("checksum")
+        # Original MIT format units are less comprehensive since they
+        # default to mV if unknown.. therefore added more default labels
+        fields.remove("units")
+
+        test_results = []
+        for field in fields:
+            # Signal value will be slightly off due to C to Python type conversion
+            if field == "p_signal":
+                true_array = np.array(record_MIT[field]).flatten()
+                pred_array = np.array(record_EDF[field]).flatten()
+                # Prevent divide by zero warning
+                for i, v in enumerate(true_array):
+                    if v == 0:
+                        true_array[i] = 1
+                        pred_array[i] = 1
+                sig_diff = np.abs((pred_array - true_array) / true_array)
+                sig_diff[sig_diff == -np.inf] = 0
+                sig_diff[sig_diff == np.inf] = 0
+                sig_diff = np.nanmean(sig_diff, 0)
+                # 5% tolerance
+                if np.max(sig_diff) <= 5:
+                    test_results.append(True)
+                else:
+                    test_results.append(False)
+            elif field == "init_value":
+                signal_diff = [
+                    abs(record_MIT[field][i] - record_EDF[field][i])
+                    for i in range(len(record_MIT[field]))
+                ]
+                if abs(max(min(signal_diff), max(signal_diff), key=abs)) <= 2:
+                    test_results.append(True)
+                else:
+                    test_results.append(False)
+            else:
+                test_results.append(record_MIT[field] == record_MIT[field])
+
+        target_results = len(fields) * [True]
+        assert np.array_equal(test_results, target_results)

tests/test_record.py

@@ -1,9 +1,9 @@
 import os
-import pdb
 import shutil
 import unittest
 
 import numpy as np
+
 import wfdb
 
 
@@ -339,110 +339,6 @@ def test_2e(self):
         sig_target = sig_target.reshape([977, 1])
         assert np.array_equal(sig, sig_target)
 
-    def test_2f(self):
-        """
-        EDF format conversion to MIT for uniform sample rates.
-
-        """
-        # Uniform sample rates
-        record_MIT = wfdb.rdrecord("sample-data/n16").__dict__
-        record_EDF = wfdb.edf2mit("sample-data/n16.edf").__dict__
-
-        fields = list(record_MIT.keys())
-        # Original MIT format method of checksum is outdated, sometimes
-        # the same value though
-        fields.remove("checksum")
-        # Original MIT format units are less comprehensive since they
-        # default to mV if unknown.. therefore added more default labels
-        fields.remove("units")
-
-        test_results = []
-        for field in fields:
-            # Signal value will be slightly off due to C to Python type conversion
-            if field == "p_signal":
-                true_array = np.array(record_MIT[field]).flatten()
-                pred_array = np.array(record_EDF[field]).flatten()
-                # Prevent divide by zero warning
-                for i, v in enumerate(true_array):
-                    if v == 0:
-                        true_array[i] = 1
-                        pred_array[i] = 1
-                sig_diff = np.abs((pred_array - true_array) / true_array)
-                sig_diff[sig_diff == -np.inf] = 0
-                sig_diff[sig_diff == np.inf] = 0
-                sig_diff = np.nanmean(sig_diff, 0)
-                # 5% tolerance
-                if np.max(sig_diff) <= 5:
-                    test_results.append(True)
-                else:
-                    test_results.append(False)
-            elif field == "init_value":
-                signal_diff = [
-                    abs(record_MIT[field][i] - record_EDF[field][i])
-                    for i in range(len(record_MIT[field]))
-                ]
-                if abs(max(min(signal_diff), max(signal_diff), key=abs)) <= 2:
-                    test_results.append(True)
-                else:
-                    test_results.append(False)
-            else:
-                test_results.append(record_MIT[field] == record_MIT[field])
-
-        target_results = len(fields) * [True]
-        assert np.array_equal(test_results, target_results)
-
-    def test_2g(self):
-        """
-        EDF format conversion to MIT for non-uniform sample rates.
-
-        """
-        # Non-uniform sample rates
-        record_MIT = wfdb.rdrecord("sample-data/wave_4").__dict__
-        record_EDF = wfdb.edf2mit("sample-data/wave_4.edf").__dict__
-
-        fields = list(record_MIT.keys())
-        # Original MIT format method of checksum is outdated, sometimes
-        # the same value though
-        fields.remove("checksum")
-        # Original MIT format units are less comprehensive since they
-        # default to mV if unknown.. therefore added more default labels
-        fields.remove("units")
-
-        test_results = []
-        for field in fields:
-            # Signal value will be slightly off due to C to Python type conversion
-            if field == "p_signal":
-                true_array = np.array(record_MIT[field]).flatten()
-                pred_array = np.array(record_EDF[field]).flatten()
-                # Prevent divide by zero warning
-                for i, v in enumerate(true_array):
-                    if v == 0:
-                        true_array[i] = 1
-                        pred_array[i] = 1
-                sig_diff = np.abs((pred_array - true_array) / true_array)
-                sig_diff[sig_diff == -np.inf] = 0
-                sig_diff[sig_diff == np.inf] = 0
-                sig_diff = np.nanmean(sig_diff, 0)
-                # 5% tolerance
-                if np.max(sig_diff) <= 5:
-                    test_results.append(True)
-                else:
-                    test_results.append(False)
-            elif field == "init_value":
-                signal_diff = [
-                    abs(record_MIT[field][i] - record_EDF[field][i])
-                    for i in range(len(record_MIT[field]))
-                ]
-                if abs(max(min(signal_diff), max(signal_diff), key=abs)) <= 2:
-                    test_results.append(True)
-                else:
-                    test_results.append(False)
-            else:
-                test_results.append(record_MIT[field] == record_MIT[field])
-
-        target_results = len(fields) * [True]
-        assert np.array_equal(test_results, target_results)
-
     # --------------------- 3. Multi-dat records --------------------- #
 
     def test_3a(self):

wfdb/__init__.py

@@ -6,34 +6,23 @@
     rdsamp,
     wrsamp,
     dl_database,
-    edf2mit,
-    mit2edf,
-    wav2mit,
-    mit2wav,
-    wfdb2mat,
-    csv2mit,
     sampfreq,
     signame,
     wfdbdesc,
     wfdbtime,
-    sigavg,
 )
 from wfdb.io.annotation import (
     Annotation,
     rdann,
     wrann,
     show_ann_labels,
     show_ann_classes,
-    ann2rr,
-    rr2ann,
-    csv2ann,
-    rdedfann,
     mrgann,
 )
 from wfdb.io.download import (
+    dl_files,
     get_dbs,
     get_record_list,
-    dl_files,
     set_db_index_url,
 )
 from wfdb.plot.plot import plot_items, plot_wfdb, plot_all_records