import numpy as np
import datajoint as dj
import nixio as nix
import os
import glob
import socket
from fishbook.backend.util import read_info_file, read_dataset_info, read_stimuli_file
from fishbook.backend.util import find_key_recursive, deep_get, find_mtags_for_tag
from fishbook.backend.util import mtag_settings_to_yaml, nix_metadata_to_yaml, mtag_features_to_yaml, progress
import uuid
import yaml
from IPython import embed
dj.config["enable_python_native_blobs"] = True
schema = dj.schema("fish_book", locals())
@schema
class Datasets(dj.Manual):
definition = """ # _Dataset
dataset_id : varchar(256)
----
data_source : varchar(512) # path to the dataset
data_host : varchar(512) # fully qualified domain name
experimenter : varchar(512)
setup : varchar(128)
recording_date : date
quality : varchar(512)
comment : varchar(1024)
duration : float
has_nix : bool
"""
@staticmethod
def get_template_tuple(id=None):
if id is not None:
d = dict((Datasets() & {"dataset_id": id}).fetch1())
return d
return dict(dataset_id=None, data_source="", data_host="", experimenter="", setup="",
recording_date=None, quality="", comment="", duration=0.0, has_nix=False)
@staticmethod
def get_nix_file(key):
dset = (Datasets() & key).fetch1()
if dset["ignore"]:
return None
file_path = os.path.join(dset["data_source"], dset["dataset_id"] + ".nix")
if not (os.path.exists(file_path)):
print("\t No nix file found for path: %s" % dset["data_source"])
return None
if not Datasets.check_file_integrity(file_path):
return None
return file_path
@staticmethod
def check_file_integrity(nix_file):
sane = True
try:
f = nix.File.open(nix_file, nix.FileMode.ReadOnly)
b = f.blocks[0]
m = b.metadata
if "Recording" not in m.sections:
Warning("\t Could not find Recording section in dataset: %s" % nix_file)
sane = False
f.close()
except ():
f = None
print("file: %s is NOT SANE!")
sane = False
return sane
@schema
class Subjects(dj.Manual):
definition = """
# Subjects
subject_id : varchar(256)
----
species : varchar(256)
"""
@staticmethod
def get_template_tuple(subject_id=None):
tup = dict(subject_id=None, species="")
if subject_id is not None:
d = dict((Subjects() & {"subject_id": subject_id}).fetch1())
return d
return tup
def make(self, key):
file_path = Datasets.get_nix_file(key)
if file_path is None:
return
nix_file = nix.File.open(file_path, nix.FileMode.ReadOnly)
m = nix_file.blocks[0].metadata
inserts = Subjects.get_template_tuple()
subj_info = m["Recording"]["Subject"]
inserts["subject_id"] = subj_info["Identifier"]
inserts["species"] = subj_info["Species"][0]
inserts["weight"] = subj_info["Weight"]
inserts["size"] = subj_info["Size"]
inserts["eod_frequency"] = np.round(subj_info["EOD Frequency"] * 10) / 10
inserts.update(key)
self.insert1(inserts, skip_duplicates=True)
nix_file.close()
#@property
#def datasets(self):
# retrun
@schema
class SubjectDatasetMap(dj.Manual):
definition = """
# SubjectDatasetMap
-> Subjects
-> Datasets
"""
@schema
class SubjectProperties(dj.Manual):
definition = """
# _SubjectProperties
id : int auto_increment
----
-> Subjects
recording_date : date
weight : float
size : float
eod_frequency : float
"""
@staticmethod
def get_template_tuple(id=None):
tup = dict(id=None, subject_id=None, recording_date=None, weight=0.0, size=0.0,
eod_frequency=0.0)
if id is not None:
return dict((SubjectProperties() & {"id": id}).fetch1())
return tup
@schema
class Cells(dj.Manual):
definition = """
# Table that stores information about recorded cells.
cell_id : varchar(256)
----
-> Subjects
cell_type : varchar(256)
firing_rate : float
structure : varchar(256)
region : varchar(256)
subregion : varchar(256)
depth : float
lateral_pos : float
transversal_section : float
"""
@staticmethod
def get_template_tuple(cell_id=None):
tup = dict(cell_id=None, subject_id=None, cell_type="", firing_rate=0.0,
depth=0.0, region="", subregion="", structure="",
lateral_pos=0.0, transversal_section=0.0)
if cell_id is not None:
d = dict((Cells() & {"cell_id": cell_id}).fetch1())
return d
return tup
@schema
class CellDatasetMap(dj.Manual):
definition = """
# Table that maps recorded cells to datasets
-> Datasets
-> Cells
"""
@schema
class Repros(dj.Manual):
definition = """
repro_id : varchar(512) # The name that was given to the RePro run by relacs
run : smallint # A counter counting the runs of the ReProp in this dataset
-> Cells #
----
repro_name : varchar(512) # The original name of the RePro itself, not any given name by user or relacs
settings : varchar(3000) # Yaml formatted string containing the repro settings (tag.metadata in case of a nix file)
start : float # The start time of the repro
duration : float # The duration of the repro
"""
@staticmethod
def get_template_tuple(repro_id=None):
tup = dict(repro_id=None, cell_id=None, run=0, repro_name="", settings=None, start=None, duration=None)
if repro_id is not None:
d = dict((Repros() & {"repro_id": repro_id}).fetch1())
return d
return tup
@schema
class Stimuli(dj.Manual):
definition = """
stimulus_id : varchar(50)
-> Repros
---
stimulus_index : int
stimulus_name : varchar(512)
mtag_id : varchar(50)
start_time : float
start_index : int
duration : float
settings : varchar(3000)
"""
@staticmethod
def get_template_tuple(stimulus_id=None):
if stimulus_id is not None:
tup = dict((Stimuli & {"stimulus_id": stimulus_id}).fetch1())
else:
tup = dict(stimulus_id=None, stimulus_index=None, stimulus_name="", start_index=0, start_time=0.0,
duration=0.0, settings=None)
return tup
def populate_datasets(data_path, update=False):
if not os.path.exists(data_path):
return False
dset_name = os.path.split(data_path)[-1]
experimenter, rec_date, quality, comment, has_nix, rec_duration, setup = read_dataset_info(os.path.join(data_path, 'info.dat'))
if not experimenter:
return False
inserts = Datasets.get_template_tuple()
inserts["dataset_id"] = dset_name
inserts["data_source"] = os.path.abspath(data_path)
inserts["data_host"] = socket.getfqdn()
inserts["experimenter"] = experimenter
inserts["recording_date"] = rec_date
inserts["quality"] = quality if not isinstance(quality, dict) else ""
inserts["comment"] = comment if not isinstance(comment, dict) else ""
inserts["duration"] = rec_duration
inserts["setup"] = setup
inserts["has_nix"] = has_nix
if len(Datasets & "dataset_id like '%s'" % inserts["dataset_id"]) > 0 and not update:
print('\t\t %s is already in database!' % dset_name)
return False
Datasets().insert1(inserts, skip_duplicates=True)
return True
def populate_subjects(data_path):
print("\tImporting subject(s) of %s" % data_path)
dset_name = os.path.split(data_path)[-1]
info_file = os.path.join(data_path, 'info.dat')
if not os.path.exists(info_file):
return None, None, False
info = read_info_file(info_file)
p = []
find_key_recursive(info, "Subject", p)
subj = {}
if len(p) > 0:
subj = deep_get(info, p)
inserts = Subjects.get_template_tuple()
subj_id = None
if "Identifier" in subj.keys():
if isinstance(subj["Identifier"], dict):
subj_id = "unspecified_" + dset_name
else:
subj_id = subj["Identifier"]
elif "Identifier" in info.keys():
if isinstance(info["Identifier"], dict):
subj_id = "unspecified_" + dset_name
else:
subj_id = info["Identifier"]
else:
subj_id = "unspecified_" + dset_name
inserts["subject_id"] = subj_id
inserts["species"] = subj["Species"]
Subjects().insert1(inserts, skip_duplicates=True)
# multi match entry
dataset = dict((Datasets() & {"dataset_id": dset_name}).fetch1())
mm = dict(dataset_id=dataset["dataset_id"], subject_id=inserts["subject_id"])
SubjectDatasetMap.insert1(mm, skip_duplicates=True)
# subject properties
props = SubjectProperties.get_template_tuple()
props["subject_id"] = inserts["subject_id"]
props["recording_date"] = dataset["recording_date"]
if "Weight" in subj.keys():
props["weight"] = np.round(float(subj["Weight"][:-1]), 1)
if "Size" in subj.keys():
props["size"] = np.round(float(subj["Size"][:-2]), 1)
if "EOD Frequency" in subj.keys():
props["eod_frequency"] = np.round(float(subj["EOD Frequency"][:-2]))
p = props.copy()
p.pop("id")
if len(SubjectProperties & p) == 0:
SubjectProperties.insert1(props, skip_duplicates=True)
def populate_cells(data_path):
print("\tImporting cell(s) of %s" % data_path)
dset_name = os.path.split(data_path)[-1]
info_file = os.path.join(data_path, 'info.dat')
if not os.path.exists(info_file):
return None, None, False
info = read_info_file(info_file)
p = []
find_key_recursive(info, "Subject", p)
subject_info = deep_get(info, p)
p = []
find_key_recursive(info, "Cell", p)
cell_info = deep_get(info, p)
p = []
res = find_key_recursive(info, "Firing Rate1", p)
if res:
firing_rate = deep_get(info, p, default=0.0)
else:
firing_rate = 0.0
if isinstance(firing_rate, str):
firing_rate = float(firing_rate[:-2])
subj_id = None
if "Identifier" in subject_info.keys():
if isinstance(subject_info["Identifier"], dict):
subj_id = "unspecified_" + dset_name
else:
subj_id = subject_info["Identifier"]
elif "Identifier" in info.keys():
if isinstance(info["Identifier"], dict):
subj_id = "unspecified_" + dset_name
else:
subj_id = info["Identifier"]
else:
subj_id = "unspecified_" + dset_name
dataset = dict((Datasets & {"dataset_id": dset_name}).fetch1())
subject = dict((Subjects & {"subject_id": subj_id}).fetch1())
dataset_id = dataset["dataset_id"]
cell_id = "-".join(dataset_id.split("-")[:4]) if len(dataset_id) > 4 else dataset_id
cell_props = Cells.get_template_tuple()
cell_props["subject_id"] = subject["subject_id"]
cell_props["cell_id"] = cell_id
cell_props["cell_type"] = cell_info["CellType"]
cell_props["firing_rate"] = firing_rate
if "Structure" in cell_info.keys():
cell_props["structure"] = cell_info["Structure"]
if "BrainRegion" in cell_info.keys():
cell_props["region"] = cell_info["BrainRegion"]
if "BrainSubRegion" in cell_info.keys():
cell_props["subregion"] = cell_info["BrainSubRegion"]
if "Depth" in cell_info.keys():
cell_props["depth"] = float(cell_info["Depth"][:-2])
if "Lateral position" in cell_info.keys():
cell_props["lateral_pos"] = float(cell_info["Lateral position"][:-2])
if "Transverse section" in cell_info.keys():
cell_props["transversal_section"] = float(cell_info["Transverse section"])
Cells.insert1(cell_props, skip_duplicates=True)
# multi match entry
mm = dict(dataset_id=dataset["dataset_id"], cell_id=cell_props["cell_id"])
CellDatasetMap.insert1(mm, skip_duplicates=True)
def scan_nix_file_for_repros(dataset):
print("\t\tscanning nix file")
cell_id = (Cells * CellDatasetMap * (Datasets & "dataset_id = '%s'" % dataset["dataset_id"])).fetch("cell_id", limit=1)[0]
nix_files = glob.glob(os.path.join(dataset["data_source"], "*.nix"))
for nf in nix_files:
if not Datasets.check_file_integrity(nf):
print("\t\tfile is not sane!!!")
continue
f = nix.File.open(nf, nix.FileMode.ReadOnly)
b = f.blocks[0]
repro_runs = [t for t in b.tags if "relacs.repro_run" in t.type]
total = len(repro_runs)
for i, t in enumerate(repro_runs):
rs = t.metadata.find_sections(lambda x: "Run" in x.props)
rs = rs[0]
if len(rs) == 0:
continue
progress(i+1, total, "Scanning repro run %s" % rs["RePro"])
rp = Repros.get_template_tuple()
rp["run"] = rs["Run"]
rp["repro_name"] = rs["RePro"]
rp["cell_id"] = cell_id
rp["repro_id"] = t.name
settings = t.metadata.find_sections(lambda x: "settings" in x.type)
if len(settings) > 0:
rp["settings"] = nix_metadata_to_yaml(settings[0])
else:
rp["settings"] = nix_metadata_to_yaml(t.metadata)
rp["start"] = t.position[0]
rp["duration"] = t.extent[0]
Repros.insert1(rp, skip_duplicates=True)
# import Stimuli
repro = dict((Repros & dict(repro_id=rp["repro_id"], cell_id=cell_id)).fetch1())
repro.pop("settings")
repro.pop("repro_name")
repro.pop("start")
repro.pop("duration")
mtags, positions = find_mtags_for_tag(b, t)
mt_settings_dict = {}
positions_dict = {}
extents_dict = {}
for j, mt in enumerate(mtags):
if mt.id in positions_dict.keys():
mt_positions = positions_dict[mt.id]
mt_extents = extents_dict[mt.id]
mdata_yaml = mt_settings_dict[mt.id]
else:
mdata_yaml = nix_metadata_to_yaml(mt.metadata)
mt_settings_dict[mt.id] = mdata_yaml
mt_positions = np.atleast_2d(mt.positions[:])
mt_extents = np.atleast_2d(mt.extents[:])
if mt.positions.shape[0] != mt_positions.shape[0]:
mt_positions = mt_positions.T
mt_extents = mt_extents.T
for p in positions[j]:
settings = mtag_features_to_yaml(mt, p, mdata_yaml)
stim_start = mt_positions[p, 0]
stim_duration = mt_extents[p, 0]
stim = Stimuli.get_template_tuple()
stim["stimulus_id"] = str(uuid.uuid1())
stim["stimulus_index"] = p
stim["start_time"] = stim_start
stim["start_index"] = -1
stim["duration"] = stim_duration
stim["settings"] = settings
stim["mtag_id"] = mt.id
stim["stimulus_name"] = mt.name
stim.update(repro)
Stimuli.insert1(stim, skip_duplicates=True)
print(" " * 120, end="\r")
print("\n")
f.close()
f = None
def scan_folder_for_repros(dataset):
print("\t\tNo nix-file, scanning directory!")
repro_settings, stim_indices = read_stimuli_file(dataset["data_source"])
repro_counts = {}
cell_id = (Cells * CellDatasetMap * (Datasets & "dataset_id = '%s'" % dataset["dataset_id"])).fetch("cell_id", limit=1)[0]
for rs, si in zip(repro_settings, stim_indices):
rp = Repros.get_template_tuple()
path = []
if not find_key_recursive(rs, "run", path):
find_key_recursive(rs, "Run", path)
if len(path) > 0:
rp["run"] = deep_get(rs, path, 0)
else:
rp["run"] = -1
path = []
if not find_key_recursive(rs, "repro", path):
find_key_recursive(rs, "RePro", path)
rp["repro_name"] = deep_get(rs, path, "None")
path = []
if rp["repro_name"] in repro_counts.keys():
repro_counts[rp["repro_name"]] += 1
else:
repro_counts[rp["repro_name"]] = 1
rp["cell_id"] = cell_id
rp["repro_id"] = rp["repro_name"] + str(repro_counts[rp["repro_name"]])
rp["start"] = 0.
rp["duration"] = 0.
rp["settings"] = yaml.dump(rs).replace("'", "")
Repros.insert1(rp, skip_duplicates=True)
# import stimuli
repro = dict((Repros & dict(repro_id=rp["repro_id"], cell_id=cell_id)).fetch1())
repro.pop("settings")
repro.pop("repro_name")
repro.pop("start")
repro.pop("duration")
total = len(si.keys())
for j, k in enumerate(si.keys()):
progress(j+1, total, "scanning repro %s" % rp["repro_name"])
s = int(si[k])
stim_start = 0.
path = []
if not find_key_recursive(rs, "duration", path):
find_key_recursive(rs, "Duration", path)
if len(path) > 0 :
stim_duration = deep_get(rs, path, None)
if "ms" in stim_duration:
stim_duration = float(stim_duration[:stim_duration.index("ms")])
else:
stim_duration = float(stim_duration[:stim_duration.index("s")])
else:
stim_duration = 0.0
stim = Stimuli.get_template_tuple()
stim["stimulus_id"] = str(uuid.uuid1())
stim["stimulus_index"] = j
stim["start_time"] = stim_start
stim["start_index"] = s
stim["duration"] = stim_duration
stim["settings"] = yaml.dump(rs).replace("'", "")
stim["mtag_id"] = ""
stim["stimulus_name"] = ""
stim.update(repro)
Stimuli.insert1(stim, skip_duplicates=True)
print(" " *120, end='\r')
#if i < len(repro_settings):
# print((" " * 150), end="\r")
def populate_repros(data_path):
print("\tImporting RePro(s) of %s" % data_path)
dset_name = os.path.split(data_path)[-1]
if len(Datasets & {"dataset_id": dset_name}) != 1:
return False
dataset = dict((Datasets & {"dataset_id": dset_name}).fetch1())
if dataset["has_nix"]:
scan_nix_file_for_repros(dataset)
else:
scan_folder_for_repros(dataset)
return True
def drop_tables():
Datasets.drop()
Subjects.drop()
def populate(datasets, update=False):
for i, d in enumerate(datasets):
print("Importing %i of %i: %s" % (i+1, len(datasets), d))
if not populate_datasets(d, update):
continue
populate_subjects(d)
populate_cells(d)
try:
populate_repros(d)
except ():
print("\t\tsomething went wrong! %s" % d)
if __name__ == "__main__":
data_dir = "/data/apteronotus"
# data_dir = "../high_freq_chirps/data"
# drop_tables()
# datasets = glob.glob("/Users/jan/zwischenlager/2012-*")2010-06-21-ac/info.dat
datasets = glob.glob(os.path.join(data_dir, '/data/apteronotus/2010-06-18*'))
populate(datasets, update=False)