def test_read_raw_on_test_data():
import glob
from rerpy.test import test_data_path
tested = 0
for rawp in glob.glob(test_data_path("erpss/*.raw")):
crwp = rawp[:-3] + "crw"
print rawp, crwp
assert_files_match(rawp, crwp)
tested += 1
# Cross-check, to make sure is actually finding the files... (bump up this
# number if you add more test files):
assert tested == 5
def test_read_raw_on_test_data():
import glob
from rerpy.test import test_data_path
tested = 0
for rawp in glob.glob(test_data_path("erpss/*.raw")):
crwp = rawp[:-3] + "crw"
print rawp, crwp
assert_files_match(rawp, crwp)
tested += 1
# Cross-check, to make sure is actually finding the files... (bump up this
# number if you add more test files):
assert tested == 5
def test_64bit_channel_names():
from rerpy.test import test_data_path
stream = open(test_data_path("erpss/two-chunks-64chan.raw"), "rb")
(_, hz, channel_names, codes, data, info) = read_raw(stream, int, True)
# "Correct" channel names as listed by headinfo(1):
assert (channel_names == [
"LOPf", "ROPf", "LMPf", "RMPf", "LTPf", "RTPf", "LLPf", "RLPf", "LPrA",
"RPrA", "LTFr", "RTFr", "LLFr", "RLFr", "LDPf", "RDPf", "LTOc", "RTOc",
"LTCe", "RTCe", "LLCe", "RLCe", "LDFr", "RDFr", "LMFr", "RMFr", "MiFo",
"MiPf", "MiFr", "A2", "LHEy", "RHEy", "LIOc", "RIOc", "LLOc", "RLOc",
"LLPP", "RLPP", "LLPa", "RLPa", "LDCe", "RDCe", "LMCe", "RMCe", "LDOc",
"RDOc", "LDPP", "RDPP", "LDPa", "RDPa", "LCer", "RCer", "LMOc", "RMOc",
"LMPP", "RMPP", "LMPa", "RMPa", "MiCe", "MiPa", "MiPP", "MiOc", "LLEy",
"RLEy"
]).all()
def test_LazyRecspan():
from nose.tools import assert_raises
from rerpy.test import test_data_path
for suffix in ["crw", "raw"]:
(fetcher, hz, channames, codes, data, info) = read_raw(
open(test_data_path("erpss/tiny-complete.%s" % (suffix, )), "rb"),
"u2", False)
# This fake recspan is chosen to cover part of the first and last
# chunks, plus the entire middle chunk. It's exactly 512 samples long.
lr = LazyRecspan(fetcher, "u2", len(channames), 128, 640)
assert_raises(IndexError, lr.get_slice, 0, 513)
for (start, stop) in [(0, 512), (10, 20), (256, 266), (500, 510),
(120, 130)]:
assert np.all(
lr.get_slice(start, stop) == data[128 + start:128 + stop])
def test_64bit_channel_names():
from rerpy.test import test_data_path
stream = open(test_data_path("erpss/two-chunks-64chan.raw"), "rb")
(_, hz, channel_names, codes, data, info) = read_raw(stream, int, True)
# "Correct" channel names as listed by headinfo(1):
assert (channel_names ==
["LOPf", "ROPf", "LMPf", "RMPf", "LTPf", "RTPf", "LLPf", "RLPf",
"LPrA", "RPrA", "LTFr", "RTFr", "LLFr", "RLFr", "LDPf", "RDPf",
"LTOc", "RTOc", "LTCe", "RTCe", "LLCe", "RLCe", "LDFr", "RDFr",
"LMFr", "RMFr", "MiFo", "MiPf", "MiFr", "A2", "LHEy", "RHEy",
"LIOc", "RIOc", "LLOc", "RLOc", "LLPP", "RLPP", "LLPa", "RLPa",
"LDCe", "RDCe", "LMCe", "RMCe", "LDOc", "RDOc", "LDPP", "RDPP",
"LDPa", "RDPa", "LCer", "RCer", "LMOc", "RMOc", "LMPP", "RMPP",
"LMPa", "RMPa", "MiCe", "MiPa", "MiPP", "MiOc", "LLEy", "RLEy"]
).all()
def test_LazyRecspan():
from nose.tools import assert_raises
from rerpy.test import test_data_path
for suffix in ["crw", "raw"]:
(fetcher, hz, channames, codes, data, info) = read_raw(
open(test_data_path("erpss/tiny-complete.%s" % (suffix,)), "rb"),
"u2", False)
# This fake recspan is chosen to cover part of the first and last
# chunks, plus the entire middle chunk. It's exactly 512 samples long.
lr = LazyRecspan(fetcher, "u2", len(channames), 128, 640)
assert_raises(IndexError, lr.get_slice, 0, 513)
for (start, stop) in [(0, 512), (10, 20), (256, 266), (500, 510),
(120, 130)]:
assert np.all(lr.get_slice(start, stop)
== data[128 + start:128 + stop])
def test_load_erpss():
from rerpy.test import test_data_path
# This crw/log file is constructed to have a few features:
# - it only has 3 records, so it's tiny
# - the first two records are in one recspan, the last is in a second, so
# we test the recspan splitting code
# - the first recspan ends in a PAUSE event, the second ends in a DELETE
# event, so we test the deleted event handling.
# There are some weird things about it too:
# - several events in the first recspan have condition 0, to test
# calibration pulse stuff. In a normal ERPSS file all events within a
# single recspan would have the same condition number.
# - most of the event codes are >32767. In a normal ERPSS file such events
# are supposed to be reserved for special stuff and deleted events, but
# it happens the file I was using as a basis violated this rule. Oh
# well.
for lazy in [False, True]:
dataset = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=lazy)
assert len(dataset) == 2
assert dataset[0].shape == (512, 32)
assert dataset[1].shape == (256, 32)
assert dataset.data_format.exact_sample_rate_hz == 250
assert dataset.data_format.units == "RAW"
assert list(dataset.data_format.channel_names) == [
"lle",
"lhz",
"MiPf",
"LLPf",
"RLPf",
"LMPf",
"RMPf",
"LDFr",
"RDFr",
"LLFr",
"RLFr",
"LMFr",
"RMFr",
"LMCe",
"RMCe",
"MiCe",
"MiPa",
"LDCe",
"RDCe",
"LDPa",
"RDPa",
"LMOc",
"RMOc",
"LLTe",
"RLTe",
"LLOc",
"RLOc",
"MiOc",
"A2",
"HEOG",
"rle",
"rhz",
]
for recspan_info in dataset.recspan_infos:
assert recspan_info["raw_file"].endswith("tiny-complete.crw")
assert recspan_info["log_file"].endswith("tiny-complete.log")
assert recspan_info["experiment"] == "brown-1"
assert recspan_info["subject"] == "Subject p3 2008-08-20"
assert recspan_info["odelay"] == 8
assert len(recspan_info["erpss_raw_header"]) == 512
assert dataset.recspan_infos[0].ticks == 512
assert dataset.recspan_infos[1].ticks == 256
assert dataset.recspan_infos[1]["deleted"]
assert len(dataset.events()) == 14
# 2 are calibration events
assert len(dataset.events("has code")) == 12
for ev in dataset.events("has code"):
assert ev["condition"] in (64, 65)
assert ev["flag"] == 0
assert not ev["flag_data_error"]
assert not ev["flag_polinv"]
assert not ev["flag_rejected"]
for ev in dataset.events("calibration_pulse"):
assert dict(ev) == {"calibration_pulse": True}
def check_ticks(query, recspan_ids, start_ticks):
events = dataset.events(query)
assert len(events) == len(recspan_ids) == len(start_ticks)
for ev, recspan_id, start_tick in zip(events, recspan_ids,
start_ticks):
assert ev.recspan_id == recspan_id
assert ev.start_tick == start_tick
assert ev.stop_tick == start_tick + 1
check_ticks("condition == 64", [0] * 8,
[21, 221, 304, 329, 379, 458, 483, 511])
check_ticks("condition == 65", [1] * 4,
[533 - 512, 733 - 512, 762 - 512, 767 - 512])
check_ticks("calibration_pulse", [0, 0], [250, 408])
# check calibration_events option
dataset2 = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=lazy,
calibration_events="condition == 65")
assert len(dataset2.events("condition == 65")) == 0
assert len(dataset2.events("condition == 0")) == 2
assert len(dataset2.events("calibration_pulse")) == 4
# check calibration
# idea: if calibration works, then the "calibration erp" will have been
# set to be the same size as whatever we told it to be.
dataset_cal = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=lazy,
calibration_events="condition == 65",
calibrate=True,
calibrate_half_width_ticks=2,
calibrate_low_cursor_time=-16,
calibrate_high_cursor_time=21,
calibrate_pulse_size=12.34,
calibrate_polarity=-1)
assert dataset_cal.data_format.units == "uV"
# -16 ms +/-2 ticks = -24 to -8 ms
low_cal = dataset_cal.rerp("calibration_pulse",
-24,
-8,
"1",
all_or_nothing=True,
overlap_correction=False)
# 21 ms rounds to 20 ms, +/-2 ticks for the window = 12 to 28 ms
high_cal = dataset_cal.rerp("calibration_pulse",
12,
28,
"1",
all_or_nothing=True,
overlap_correction=False)
low = low_cal.betas["Intercept"].mean(axis=0)
high = high_cal.betas["Intercept"].mean(axis=0)
assert np.allclose(high - low, -1 * 12.34)
# check that we can load from file handles (not sure if anyone cares but
# hey you never know...)
crw = open(test_data_path("erpss/tiny-complete.crw"), "rb")
log = open(test_data_path("erpss/tiny-complete.log"), "rb")
assert len(load_erpss(crw, log, lazy=lazy)) == 2
# check that code/raw mismatch is detected
from nose.tools import assert_raises
for bad in ["bad-code", "bad-tick", "bad-tick2"]:
assert_raises(ValueError,
load_erpss,
test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.%s.log" %
(bad, )),
lazy=lazy)
# But if the only mismatch is an event that is "deleted" (sign bit
# set) in the log file, but not in the raw file, then that is okay:
load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.code-deleted.log"),
lazy=lazy)
# Compare lazy to eager directly
eager = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=False)
lazy = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=True)
from pandas.util.testing import assert_frame_equal
assert len(eager) == len(lazy)
for eager_recspan, lazy_recspan in zip(eager, lazy):
assert_frame_equal(eager_recspan, lazy_recspan)
def test_load_erpss():
from rerpy.test import test_data_path
# This crw/log file is constructed to have a few features:
# - it only has 3 records, so it's tiny
# - the first two records are in one recspan, the last is in a second, so
# we test the recspan splitting code
# - the first recspan ends in a PAUSE event, the second ends in a DELETE
# event, so we test the deleted event handling.
# There are some weird things about it too:
# - several events in the first recspan have condition 0, to test
# calibration pulse stuff. In a normal ERPSS file all events within a
# single recspan would have the same condition number.
# - most of the event codes are >32767. In a normal ERPSS file such events
# are supposed to be reserved for special stuff and deleted events, but
# it happens the file I was using as a basis violated this rule. Oh
# well.
for lazy in [False, True]:
dataset = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=lazy)
assert len(dataset) == 2
assert dataset[0].shape == (512, 32)
assert dataset[1].shape == (256, 32)
assert dataset.data_format.exact_sample_rate_hz == 250
assert dataset.data_format.units == "RAW"
assert list(dataset.data_format.channel_names) == [
"lle", "lhz", "MiPf", "LLPf", "RLPf", "LMPf", "RMPf", "LDFr", "RDFr",
"LLFr", "RLFr", "LMFr", "RMFr", "LMCe", "RMCe", "MiCe", "MiPa", "LDCe",
"RDCe", "LDPa", "RDPa", "LMOc", "RMOc", "LLTe", "RLTe", "LLOc", "RLOc",
"MiOc", "A2", "HEOG", "rle", "rhz",
]
for recspan_info in dataset.recspan_infos:
assert recspan_info["raw_file"].endswith("tiny-complete.crw")
assert recspan_info["log_file"].endswith("tiny-complete.log")
assert recspan_info["experiment"] == "brown-1"
assert recspan_info["subject"] == "Subject p3 2008-08-20"
assert recspan_info["odelay"] == 8
assert len(recspan_info["erpss_raw_header"]) == 512
assert dataset.recspan_infos[0].ticks == 512
assert dataset.recspan_infos[1].ticks == 256
assert dataset.recspan_infos[1]["deleted"]
assert len(dataset.events()) == 14
# 2 are calibration events
assert len(dataset.events("has code")) == 12
for ev in dataset.events("has code"):
assert ev["condition"] in (64, 65)
assert ev["flag"] == 0
assert not ev["flag_data_error"]
assert not ev["flag_polinv"]
assert not ev["flag_rejected"]
for ev in dataset.events("calibration_pulse"):
assert dict(ev) == {"calibration_pulse": True}
def check_ticks(query, recspan_ids, start_ticks):
events = dataset.events(query)
assert len(events) == len(recspan_ids) == len(start_ticks)
for ev, recspan_id, start_tick in zip(events, recspan_ids, start_ticks):
assert ev.recspan_id == recspan_id
assert ev.start_tick == start_tick
assert ev.stop_tick == start_tick + 1
check_ticks("condition == 64",
[0] * 8, [21, 221, 304, 329, 379, 458, 483, 511])
check_ticks("condition == 65",
[1] * 4,
[533 - 512, 733 - 512, 762 - 512, 767 - 512])
check_ticks("calibration_pulse", [0, 0], [250, 408])
# check calibration_events option
dataset2 = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=lazy,
calibration_events="condition == 65")
assert len(dataset2.events("condition == 65")) == 0
assert len(dataset2.events("condition == 0")) == 2
assert len(dataset2.events("calibration_pulse")) == 4
# check calibration
# idea: if calibration works, then the "calibration erp" will have been
# set to be the same size as whatever we told it to be.
dataset_cal = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=lazy,
calibration_events="condition == 65",
calibrate=True,
calibrate_half_width_ticks=2,
calibrate_low_cursor_time=-16,
calibrate_high_cursor_time=21,
calibrate_pulse_size=12.34,
calibrate_polarity=-1)
assert dataset_cal.data_format.units == "uV"
# -16 ms +/-2 ticks = -24 to -8 ms
low_cal = dataset_cal.rerp("calibration_pulse", -24, -8, "1",
all_or_nothing=True,
overlap_correction=False)
# 21 ms rounds to 20 ms, +/-2 ticks for the window = 12 to 28 ms
high_cal = dataset_cal.rerp("calibration_pulse", 12, 28, "1",
all_or_nothing=True,
overlap_correction=False)
low = low_cal.betas["Intercept"].mean(axis=0)
high = high_cal.betas["Intercept"].mean(axis=0)
assert np.allclose(high - low, -1 * 12.34)
# check that we can load from file handles (not sure if anyone cares but
# hey you never know...)
crw = open(test_data_path("erpss/tiny-complete.crw"), "rb")
log = open(test_data_path("erpss/tiny-complete.log"), "rb")
assert len(load_erpss(crw, log, lazy=lazy)) == 2
# check that code/raw mismatch is detected
from nose.tools import assert_raises
for bad in ["bad-code", "bad-tick", "bad-tick2"]:
assert_raises(ValueError,
load_erpss,
test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.%s.log" % (bad,)),
lazy=lazy)
# But if the only mismatch is an event that is "deleted" (sign bit
# set) in the log file, but not in the raw file, then that is okay:
load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.code-deleted.log"),
lazy=lazy)
# Compare lazy to eager directly
eager = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=False)
lazy = load_erpss(test_data_path("erpss/tiny-complete.crw"),
test_data_path("erpss/tiny-complete.log"),
lazy=True)
from pandas.util.testing import assert_frame_equal
assert len(eager) == len(lazy)
for eager_recspan, lazy_recspan in zip(eager, lazy):
assert_frame_equal(eager_recspan, lazy_recspan)