Source code for pyiron.gaussian.gaussian

# coding: utf-8
# Copyright (c) Max-Planck-Institut für Eisenforschung GmbH - Computational Materials Design (CM) Department
# Distributed under the terms of "New BSD License", see the LICENSE file.

import os,subprocess,re,pandas
import numpy as np
import matplotlib.pyplot as pt

from pyiron.dft.job.generic import GenericDFTJob
from pyiron.base.generic.parameters import GenericParameters
from pyiron.atomistics.structure.atoms import Atoms

try:
    from molmod.io.fchk import FCHKFile
    from molmod.units import amu,angstrom,electronvolt,centimeter,kcalmol
    from molmod.constants import lightspeed
    from molmod.periodic import periodic
    import tamkin
except ImportError:
    pass


__author__ = "Jan Janssen, Sander Borgmans"
__copyright__ = "Copyright 2019, Max-Planck-Institut für Eisenforschung GmbH - " \
                "- Computational Materials Design (CM) Department"
__version__ = "1.0"
__maintainer__ = ""
__email__ = ""
__status__ = "trial"
__date__ = "Aug 27, 2019"


[docs]class Gaussian(GenericDFTJob): def __init__(self, project, job_name): super(Gaussian, self).__init__(project, job_name) self.__name__ = "Gaussian" self._executable_activate(enforce=True) self.input = GaussianInput()
[docs] def write_input(self): input_dict = {'mem': self.server.memory_limit, 'cores': self.server.cores, 'verbosity': self.input['verbosity'], 'lot': self.input['lot'], 'basis_set': self.input['basis_set'], 'jobtype' : self.input['jobtype'], 'settings' : self.input['settings'], 'title' : self.input['title'], 'spin_mult': self.input['spin_mult'], 'charge': self.input['charge'], 'bsse_idx': self.input['bsse_idx'], 'symbols': self.structure.get_chemical_symbols().tolist(), 'pos': self.structure.positions } write_input(input_dict=input_dict, working_directory=self.working_directory)
[docs] def collect_output(self): output_dict = collect_output(output_file=os.path.join(self.working_directory, 'input.fchk')) with self.project_hdf5.open("output") as hdf5_output: for k, v in output_dict.items(): hdf5_output[k] = v
[docs] def to_hdf(self, hdf=None, group_name=None): super(Gaussian, self).to_hdf(hdf=hdf, group_name=group_name) with self.project_hdf5.open("input") as hdf5_input: self.structure.to_hdf(hdf5_input) self.input.to_hdf(hdf5_input)
[docs] def from_hdf(self, hdf=None, group_name=None): super(Gaussian, self).from_hdf(hdf=hdf, group_name=group_name) with self.project_hdf5.open("input") as hdf5_input: self.input.from_hdf(hdf5_input) self.structure = Atoms().from_hdf(hdf5_input)
[docs] def log(self): with open(os.path.join(self.working_directory, 'input.log')) as f: print(f.read())
[docs] def calc_minimize(self, electronic_steps=None, ionic_steps=None, algorithm=None, ionic_forces=None): ''' Function to setup the hamiltonian to perform ionic relaxations using DFT. The convergence goal can be set using either the iconic_energy as an limit for fluctuations in energy or the iconic_forces. **Arguments** algorithm: SCF algorithm electronic_steps (int): maximum number of electronic steps per electronic convergence ionic_steps (int): maximum number of ionic steps ionic_forces ('tight' or 'verytight'): convergence criterium for Berny opt (optional) ''' settings = {} opt_settings = [] if electronic_steps is not None: if not 'SCF' in settings: settings['SCF'] = [] settings['SCF'].append("MaxCycle={}".format(electronic_steps)) if ionic_steps is not None: opt_settings.append("MaxCycles={}".format(ionic_steps)) if algorithm is not None: if not 'SCF' in settings: settings['SCF'] = [] settings['SCF'].append(algorithm) if ionic_forces is not None: assert isinstance(ionic_forces,str) opt_settings.append(ionic_forces) self.input['jobtype'] = 'opt' + '({})'.format(",".join(opt_settings))*(len(opt_settings)>0) if not isinstance(self.input['settings'],dict): self.input['settings'] = settings else: self.input['settings'].update(settings) super(Gaussian, self).calc_minimize( electronic_steps=electronic_steps, ionic_steps=ionic_steps, algorithm=algorithm, ionic_forces=ionic_forces )
[docs] def calc_static(self, electronic_steps=None, algorithm=None): ''' Function to setup the hamiltonian to perform static SCF DFT runs **Arguments** algorithm (str): SCF algorithm electronic_steps (int): maximum number of electronic steps, which can be used to achieve convergence ''' settings = {} if electronic_steps is not None: if not 'SCF' in settings: settings['SCF'] = [] settings['SCF'].append("MaxCycle={}".format(electronic_steps)) if algorithm is not None: if not 'SCF' in settings: settings['SCF'] = [] settings['SCF'].append(algorithm) self.input['jobtype'] = 'sp' if not isinstance(self.input['settings'],dict): self.input['settings'] = settings else: self.input['settings'].update(settings) super(Gaussian, self).calc_static( electronic_steps=electronic_steps, algorithm=algorithm )
[docs] def calc_md(self, temperature=None, n_ionic_steps=1000, time_step=None, n_print=100): raise NotImplementedError("calc_md() not implemented in Gaussian.")
[docs] def print_MO(self): ''' Print a list of the MO's with the corresponding orbital energy and occupation. ''' n_MO = self.get('output/structure/dft/scf_density').shape[0] for n,index in enumerate(range(n_MO)): # print orbital information occ_alpha = int(self.get('output/structure/dft/n_alpha_electrons') > index) occ_beta = int(self.get('output/structure/dft/n_beta_electrons') > index) if self.get('output/structure/dft/beta_orbital_e') is None: orbital_energy = self.get('output/structure/dft/alpha_orbital_e')[index] print("#{}: \t Orbital energy = {:>10.5f} \t Occ. = {}".format(n,orbital_energy,occ_alpha+occ_beta)) else: orbital_energy = [self.get('output/structure/dft/alpha_orbital_e')[index],self.get('output/structure/dft/beta_orbital_e')[index]] print("#{}: \t Orbital energies (alpha,beta) = {:>10.5f},{:>10.5f} \t Occ. = {},{}".format(n,orbital_energy[0],orbital_energy[1],occ_alpha,occ_beta))
[docs] def visualize_MO(self,index,particle_size=0.5,show_bonds=True): ''' Visualize the MO identified by its index. **Arguments** index index of the MO, as listed by print_MO() particle_size size of the atoms for visualization, lower value if orbital is too small to see show_bonds connect atoms or not **Notes** This function should always be accompanied with the following commands (in a separate cell) view[1].update_surface(isolevel=1, color='blue', opacity=.3) view[2].update_surface(isolevel=-1, color='red', opacity=.3) This makes sure that the bonding and non-bonding MO's are plotted and makes them transparent ''' n_MO = self.get('output/structure/dft/scf_density').shape[0] assert index >= 0 and index < n_MO assert len(self.get('output/structure/numbers')) < 50 # check whether structure does not become too large for interactive calculation of cube file # print orbital information occ_alpha = int(self.get('output/structure/dft/n_alpha_electrons') > index) occ_beta = int(self.get('output/structure/dft/n_beta_electrons') > index) if self.get('output/structure/dft/beta_orbital_e') is None: orbital_energy = self.get('output/structure/dft/alpha_orbital_e')[index] print("Orbital energy = {:>10.5f} \t Occ. = {}".format(orbital_energy,occ_alpha+occ_beta)) else: orbital_energy = [self.get('output/structure/dft/alpha_orbital_e')[index],self.get('output/structure/dft/beta_orbital_e')[index]] print("Orbital energies (alpha,beta) = {:>10.5f},{:>10.5f} \t Occ. = {},{}".format(orbital_energy[0],orbital_energy[1],occ_alpha,occ_beta)) # make cube file path = self.path+'_hdf5/'+self.name+'/input' out = subprocess.check_output( "ml load Gaussian/g16_E.01-intel-2019a;module use /apps/gent/CO7/haswell-ib/modules/all; cubegen 1 MO={} {}.fchk {}.cube".format(index+1,path,path), stderr=subprocess.STDOUT, universal_newlines=True, shell=True, ) # visualize cube file try: import nglview except ImportError: raise ImportError("The animate_nma_mode() function requires the package nglview to be installed") atom_numbers = [] atom_positions = [] with open('{}.cube'.format(path),'r') as f: for i in range(2): f.readline() n_atoms = int(f.readline().split()[0][1:]) for i in range(3): f.readline() for n in range(n_atoms): line = f.readline().split() atom_numbers.append(int(line[0])) atom_positions.append(np.array([float(m) for m in line[2:]])/angstrom) structure = Atoms(numbers=np.array(atom_numbers),positions=atom_positions) view = nglview.show_ase(structure) if not show_bonds: view.add_spacefill(radius_type='vdw', scale=0.5, radius=particle_size) view.remove_ball_and_stick() else: view.add_ball_and_stick() view.add_component('{}.cube'.format(path)) view.add_component('{}.cube'.format(path)) return view
[docs] def read_NMA(self): ''' Reads the NMA output from the Gaussian .log file. Returns: IR frequencies, intensities and corresponding eigenvectors (modes). ''' # Read number of atoms nrat = len(self.get('output/structure/numbers')) # Read IR frequencies and intensities from log file low_freqs = [] freqs = [] ints = [] modes = [[] for i in range(nrat)] path = self.path+'_hdf5/'+self.name+'/input.log' with open(path,'r') as f: lines = f.readlines() # Assert normal termination assert "Normal termination of Gaussian" in lines[-1] # Find zero frequencies for n in range(len(lines)): line = lines[n] if 'Low frequencies' in line: low_freqs += [float(i) for i in line[20:].split()] if 'Frequencies --' in line: freqs += [float(i) for i in line[15:].split()] if 'IR Inten --' in line: ints += [float(i) for i in line[15:].split()] if 'Atom AN X Y Z' in line: for m in range(nrat): modes[m] += [float(i) for i in lines[n+m+1][10:].split()] nma_zeros = 3*nrat-len(freqs) freq_array = np.zeros(3*nrat) freq_array[:nma_zeros] = np.array(low_freqs[:nma_zeros]) freq_array[nma_zeros:] = np.array(freqs) freqs = freq_array * (lightspeed/centimeter) # put into atomic units ints = np.array(ints) modes = np.array(modes).reshape(len(ints),nrat,3) return freqs,ints,modes
[docs] def bsse_to_pandas(self): ''' Convert bsse output of all frames to a pandas Dataframe object. Returns: pandas.Dataframe: output as dataframe ''' assert 'counterpoise' in [k.lower() for k in self.input['settings'].keys()] # check if there was a bsse calculation tmp = {} with self.project_hdf5.open('output/structure/bsse') as hdf: for key in hdf.list_nodes(): tmp[key] = hdf[key] if isinstance(hdf[key],np.ndarray) else [hdf[key]] df = pandas.DataFrame(tmp) return df
[docs]class GaussianInput(GenericParameters): def __init__(self, input_file_name=None): super(GaussianInput, self).__init__(input_file_name=input_file_name, table_name="input_inp", comment_char="#")
[docs] def load_default(self): ''' Loading the default settings for the input file. ''' input_str = """\ lot HF basis_set 6-311G(d,p) spin_mult 1 charge 0 """ self.load_string(input_str)
[docs]def write_input(input_dict,working_directory='.'): # Comments can be written with ! in Gaussian # Load dictionary lot = input_dict['lot'] basis_set = input_dict['basis_set'] spin_mult = input_dict['spin_mult'] # 2S+1 charge = input_dict['charge'] symbols = input_dict['symbols'] pos = input_dict['pos'] assert pos.shape[0] == len(symbols) # Optional elements if not input_dict['mem'] is None: mem = input_dict['mem'] + 'B' * (input_dict['mem'][-1]!='B') # check if string ends in bytes # convert pmem to mem cores = input_dict['cores'] nmem = str(int(re.findall("\d+", mem)[0]) * cores) mem_unit = re.findall("[a-zA-Z]+", mem)[0] mem = nmem+mem_unit else: mem = "800MB" # default allocation if not input_dict['jobtype'] is None: jobtype = input_dict['jobtype'] else: jobtype = "" # corresponds to sp if not input_dict['title'] is None: title = input_dict['title'] else: title = "no title" if not input_dict['settings'] is None: settings = input_dict['settings'] # dictionary {key: [options]} else: settings = {} verbosity_dict={'low':'t','normal':'n','high':'p'} if not input_dict['verbosity'] is None: verbosity = input_dict['verbosity'] if verbosity in verbosity_dict: verbosity = verbosity_dict[verbosity] else: verbosity='n' if 'Counterpoise' in settings.keys(): if input_dict['bsse_idx'] is None or not len(input_dict['bsse_idx'])==len(pos) : # check if all elements are present for a BSSE calculation raise ValueError('The Counterpoise setting requires a valid bsse_idx array') # Check bsse idx (should start from 1 for Gaussian) input_dict['bsse_idx'] = [k - min(input_dict['bsse_idx']) + 1 for k in input_dict['bsse_idx']] # Check if it only contains conseqcutive numbers (sum of set should be n*(n+1)/2) assert sum(set(input_dict['bsse_idx'])) == (max(input_dict['bsse_idx'])*(max(input_dict['bsse_idx']) + 1))/2 # Parse settings settings_string = "" for key,valuelst in settings.items(): if not isinstance(valuelst, list): valuelst = [valuelst] option = key + "({}) ".format(",".join(valuelst))*(len(valuelst)>0) settings_string += option # Write to file route_section = "#{} {}/{} {} {}\n\n".format(verbosity,lot,basis_set,jobtype,settings_string) with open(os.path.join(working_directory, 'input.com'), 'w') as f: f.write("%mem={}\n".format(mem)) f.write("%chk=input.chk\n") f.write(route_section) f.write("{}\n\n".format(title)) if not 'Counterpoise' in settings.keys(): f.write("{} {}\n".format(charge,spin_mult)) for n,p in enumerate(pos): f.write(" {}\t{: 1.6f}\t{: 1.6f}\t{: 1.6f}\n".format(symbols[n],p[0],p[1],p[2])) f.write('\n\n') # don't know whether this is still necessary in G16 else: if isinstance(charge,list) and isinstance(spin_mult,list): # for BSSE it is possible to define charge and multiplicity for the fragments separately f.write(" ".join(["{},{}".format(charge[idx],spin_mult[idx]) for idx in range(int(settings['Counterpoise']))])) # first couple is for full system, then every fragment separately else: f.write("{} {}\n".format(charge,spin_mult)) for n,p in enumerate(pos): f.write(" {}(Fragment={})\t{: 1.6f}\t{: 1.6f}\t{: 1.6f}\n".format(symbols[n],input_dict['bsse_idx'][n],p[0],p[1],p[2])) f.write('\n\n') # don't know whether this is still necessary in G16
# we could use theochem iodata, should be more robust than molmod.io # but we require the latest iodata for this, not the conda version
[docs]def fchk2dict(fchk): # probably still some data missing # check job type, for now implement basics (SP=single point, FOpt = full opt, Freq = frequency calculation) if not fchk.command.lower() in ['sp','fopt','freq']: raise NotImplementedError # Basic information fchkdict = {} fchkdict['jobtype'] = fchk.command.lower() fchkdict['lot'] = fchk.lot fchkdict['basis_set'] = fchk.basis fchkdict['structure/numbers'] = fchk.fields.get('Atomic numbers') fchkdict['structure/masses'] = fchk.fields.get('Real atomic weights')*amu fchkdict['structure/charges'] = fchk.fields.get('Mulliken Charges') fchkdict['structure/dipole'] = fchk.fields.get('Dipole Moment') fchkdict['structure/dft/n_electrons'] = fchk.fields.get('Number of electrons') fchkdict['structure/dft/n_alpha_electrons'] = fchk.fields.get('Number of alpha electrons') fchkdict['structure/dft/n_beta_electrons'] = fchk.fields.get('Number of beta electrons') fchkdict['structure/dft/n_basis_functions'] = fchk.fields.get('Number of basis functions') # Orbital information fchkdict['structure/dft/alpha_orbital_e'] = fchk.fields.get('Alpha Orbital Energies') fchkdict['structure/dft/beta_orbital_e'] = fchk.fields.get('Beta Orbital Energies') # Densities fchkdict['structure/dft/scf_density'] = _triangle_to_dense(fchk.fields.get('Total SCF Density')) fchkdict['structure/dft/spin_scf_density'] = _triangle_to_dense(fchk.fields.get('Spin SCF Density')) if fchk.lot.upper() in ['MP2', 'MP3', 'CC', 'CI']: # only one of the lots should be present, hence using the same key fchkdict['structure/dft/post_scf_density'] = _triangle_to_dense(fchk.fields.get('Total {} Density'.format(fchk.lot))) fchkdict['structure/dft/post_spin_scf_density'] = _triangle_to_dense(fchk.fields.get('Spin {} Density'.format(fchk.lot))) # Specific job information if fchkdict['jobtype'] == 'fopt': if len(fchk.get_optimization_coordinates().shape) == 3: fchkdict['structure/positions'] = fchk.get_optimization_coordinates()[-1]/angstrom else: fchkdict['structure/positions'] = fchk.get_optimization_coordinates()/angstrom fchkdict['generic/positions'] = fchk.get_optimization_coordinates()/angstrom fchkdict['generic/energy_tot'] = fchk.get_optimization_energies()/electronvolt fchkdict['generic/forces'] = fchk.get_optimization_gradients()/(electronvolt/angstrom) * -1 if fchkdict['jobtype'] == 'freq': fchkdict['structure/positions'] = fchk.fields.get('Current cartesian coordinates').reshape([1,-1, 3])/angstrom fchkdict['generic/positions'] = fchk.fields.get('Current cartesian coordinates').reshape([1,-1, 3])/angstrom fchkdict['generic/forces'] = fchk.fields.get('Cartesian Gradient').reshape([-1, 3])/(electronvolt/angstrom) *-1 fchkdict['generic/hessian'] = fchk.get_hessian()/(electronvolt/angstrom**2) fchkdict['generic/energy_tot'] = fchk.fields.get('Total Energy')/electronvolt if fchkdict['jobtype'] == 'sp': fchkdict['structure/positions'] = fchk.fields.get('Current cartesian coordinates').reshape([1,-1, 3])/angstrom fchkdict['generic/positions'] = fchk.fields.get('Current cartesian coordinates').reshape([1,-1, 3])/angstrom fchkdict['generic/energy_tot'] = fchk.fields.get('Total Energy')/electronvolt return fchkdict
[docs]def get_bsse_array(line,it): numeric_const_pattern = '[-+]? (?: (?: \d* \. \d+ ) | (?: \d+ \.? ) )(?: [Ee] [+-]? \d+ ) ?' rx = re.compile(numeric_const_pattern, re.VERBOSE) cE_corr = float(rx.findall(line)[0]) * kcalmol/electronvolt line = next(it) # go to next line cE_raw = float(rx.findall(line)[0]) * kcalmol/electronvolt line = next(it) # go to next line sum_fragments = float(rx.findall(line)[0])/electronvolt line = next(it) # go to next line bsse_corr = float(rx.findall(line)[0])/electronvolt line = next(it) # go to next line E_tot_corr = float(rx.findall(line)[0])/electronvolt return E_tot_corr,bsse_corr,sum_fragments,cE_raw,cE_corr
[docs]def read_bsse(output_file,output_dict): # Check whether the route section contains the Counterpoise setting (if fchk module is update, route section can be loaded from dict) cp = False with open(output_file,'r') as f: line = f.readline() while line: if 'route' in line.lower(): if 'counterpoise' in f.readline().lower(): # read next line cp = True break line = f.readline() if cp: # the log file has the same path and name as the output file aside from the file extension log_file = output_file[:output_file.rfind('.')] + '.log' frames = 1 if isinstance(output_dict['generic/energy_tot'],float) else len(output_dict['generic/energy_tot']) output_dict['structure/bsse/energy_tot_corrected'] = np.zeros(frames) output_dict['structure/bsse/bsse_correction'] = np.zeros(frames) output_dict['structure/bsse/sum_of_fragments'] = np.zeros(frames) output_dict['structure/bsse/complexation_energy_raw'] = np.zeros(frames) output_dict['structure/bsse/complexation_energy_corrected'] = np.zeros(frames) it = _reverse_readline(log_file) line = next(it) for i in range(frames): found = False while not found: line = next(it) if 'complexation energy' in line: E_tot_corr,bsse_corr,sum_fragments,cE_raw,cE_corr = get_bsse_array(line,it) output_dict['structure/bsse/energy_tot_corrected'][i] = E_tot_corr output_dict['structure/bsse/bsse_correction'][i] = bsse_corr output_dict['structure/bsse/sum_of_fragments'][i] = sum_fragments output_dict['structure/bsse/complexation_energy_raw'][i] = cE_raw output_dict['structure/bsse/complexation_energy_corrected'][i] = cE_corr found = True if frames==1: output_dict['structure/bsse/energy_tot_corrected'] = output_dict['structure/bsse/energy_tot_corrected'][0] output_dict['structure/bsse/bsse_correction'] = output_dict['structure/bsse/bsse_correction'][0] output_dict['structure/bsse/sum_of_fragments'] = output_dict['structure/bsse/sum_of_fragments'][0] output_dict['structure/bsse/complexation_energy_raw'] = output_dict['structure/bsse/complexation_energy_raw'][0] output_dict['structure/bsse/complexation_energy_corrected'] = output_dict['structure/bsse/complexation_energy_corrected'][0] else: # flip array sequence output_dict['structure/bsse/energy_tot_corrected'] = output_dict['structure/bsse/energy_tot_corrected'][::-1] output_dict['structure/bsse/bsse_correction'] = output_dict['structure/bsse/bsse_correction'][::-1] output_dict['structure/bsse/sum_of_fragments'] = output_dict['structure/bsse/sum_of_fragments'][::-1] output_dict['structure/bsse/complexation_energy_raw'] = output_dict['structure/bsse/complexation_energy_raw'][::-1] output_dict['structure/bsse/complexation_energy_corrected'] = output_dict['structure/bsse/complexation_energy_corrected'][::-1]
[docs]def read_EmpiricalDispersion(output_file,output_dict): # Get dispersion term from log file if it is there # dispersion term is not retrieved from gaussian output in fchk disp = None with open(output_file,'r') as f: while True: line = f.readline() if 'Route' in line: line = f.readline() if 'EmpiricalDispersion' in line: idx = line.find('EmpiricalDispersion') if 'GD3' in line[idx:]: search_term = 'Grimme-D3 Dispersion energy=' else: raise NotImplementedError else: return break # the log file has the same path and name as the output file aside from the file extension log_file = output_file[:output_file.rfind('.')] + '.log' it = _reverse_readline(log_file) while True: line = next(it) if search_term in line: disp = float(line[38:-9])/electronvolt # could be changed when new search terms are implemented break output_dict['generic/energy_tot'] += disp
[docs]def collect_output(output_file): # Read output fchk = FCHKFile(output_file) # Translate to dict output_dict = fchk2dict(fchk) # Read BSSE output if it is present read_bsse(output_file,output_dict) # Correct energy if empirical dispersion contribution is present read_EmpiricalDispersion(output_file,output_dict) return output_dict
# function from theochem iodata def _triangle_to_dense(triangle): '''Convert a symmetric matrix in triangular storage to a dense square matrix. Parameters ---------- triangle A row vector containing all the unique matrix elements of symmetric matrix. (Either the lower-triangular part in row major-order or the upper-triangular part in column-major order.) Returns ------- ndarray a square symmetric matrix. ''' if triangle is None: return None nrow = int(np.round((np.sqrt(1 + 8 * len(triangle)) - 1) / 2)) result = np.zeros((nrow, nrow)) begin = 0 for irow in range(nrow): end = begin + irow + 1 result[irow, :irow + 1] = triangle[begin:end] result[:irow + 1, irow] = triangle[begin:end] begin = end return result def _reverse_readline(filename, buf_size=8192): '''A generator that returns the lines of a file in reverse order''' '''https://stackoverflow.com/questions/2301789/read-a-file-in-reverse-order-using-python''' with open(filename) as fh: segment = None offset = 0 fh.seek(0, os.SEEK_END) file_size = remaining_size = fh.tell() while remaining_size > 0: offset = min(file_size, offset + buf_size) fh.seek(file_size - offset) buffer = fh.read(min(remaining_size, buf_size)) remaining_size -= buf_size lines = buffer.split('\n') # The first line of the buffer is probably not a complete line so # we'll save it and append it to the last line of the next buffer # we read if segment is not None: # If the previous chunk starts right from the beginning of line # do not concat the segment to the last line of new chunk. # Instead, yield the segment first if buffer[-1] != '\n': lines[-1] += segment else: yield segment segment = lines[0] for index in range(len(lines) - 1, 0, -1): if lines[index]: yield lines[index] # Don't yield None if the file was empty if segment is not None: yield segment