Mercurial > repos > pieterlukasse > prims_metabolomics
view match_library.py @ 62:9bd2597c8851 default tip
r
author | pieter.lukasse@wur.nl |
---|---|
date | Fri, 06 Feb 2015 15:49:26 +0100 |
parents | e67149fbff20 |
children |
line wrap: on
line source
''' Containing functions are called from Galaxy to populate lists/checkboxes with selectable items ''' import csv import glob import os __author__ = "Marcel Kempenaar" __contact__ = "brs@nbic.nl" __copyright__ = "Copyright, 2012, Netherlands Bioinformatics Centre" __license__ = "MIT" def get_column_type(library_file): ''' Returns a Galaxy formatted list of tuples containing all possibilities for the GC-column types. Used by the library_lookup.xml tool @param library_file: given library file from which the list of GC-column types is extracted ''' if library_file == "": galaxy_output = [("", "", False)] else: (data, header) = read_library(library_file) if 'columntype' not in header: raise IOError('Missing columns in ', library_file) # Filter data on column type column_type = header.index("columntype") amounts_in_list_dict = count_occurrence([row[column_type] for row in data]) galaxy_output = [(str(a) + "(" + str(b) + ")", a, False) for a, b in amounts_in_list_dict.items()] return(galaxy_output) def filter_column(library_file, column_type_name): ''' Filters the Retention Index database on column type @param library_file: file containing the database @param column_type_name: column type to filter on ''' if library_file == "": galaxy_output = [("", "", False)] else: (data, header) = read_library(library_file) if ('columntype' not in header or 'columnphasetype' not in header): raise IOError('Missing columns in ', library_file) column_type = header.index("columntype") statphase = header.index("columnphasetype") # Filter data on colunn type name statphase_list = [line[statphase] for line in data if line[column_type] == column_type_name] amounts_in_list_dict = count_occurrence(statphase_list) galaxy_output = [(str(a) + "(" + str(b) + ")", a, False)for a, b in amounts_in_list_dict.items()] return(sorted(galaxy_output)) def filter_column2(library_file, column_type_name, statphase): ''' Filters the Retention Index database on column type @param library_file: file containing the database @param column_type_name: column type to filter on @param statphase: stationary phase of the column to filter on ''' if library_file == "": galaxy_output = [("", "", False)] else: (data, header) = read_library(library_file) if ('columntype' not in header or 'columnphasetype' not in header or 'columnname' not in header): raise IOError('Missing columns in ', library_file) column_type_column = header.index("columntype") statphase_column = header.index("columnphasetype") column_name_column = header.index("columnname") # Filter data on given column type name and stationary phase statphase_list = [line[column_name_column] for line in data if line[column_type_column] == column_type_name and line[statphase_column] == statphase] amounts_in_list_dict = count_occurrence(statphase_list) galaxy_output = [(str(a) + "(" + str(b) + ")", a, False)for a, b in amounts_in_list_dict.items()] return(sorted(galaxy_output)) def read_library(filename): ''' Reads a CSV file and returns its contents and a normalized header @param filename: file to read ''' data = list(csv.reader(open(filename, 'rU'), delimiter='\t')) header_clean = [i.lower().strip().replace(".", "").replace("%", "") for i in data.pop(0)] return(data, header_clean) def get_directory_files(dir_name): ''' Reads the directory and returns the list of .txt files found as a dictionary with file name and full path so that it can fill a Galaxy drop-down combo box. ''' files = glob.glob(dir_name + "/*.*") if len(files) == 0: # Configuration error: no library files found in <galaxy-home-dir>/" + dir_name : galaxy_output = [("Configuration error: expected file not found in <galaxy-home-dir>/" + dir_name, "", False)] else: galaxy_output = [(str(get_file_name_no_ext(file_name)), str(os.path.abspath(file_name)), False) for file_name in files] return(galaxy_output) def get_file_name_no_ext(full_name): ''' returns just the last part of the name ''' simple_name = os.path.basename(full_name) base, ext = os.path.splitext(simple_name) return base def count_occurrence(data_list): ''' Counts occurrences in a list and returns a dict with item:occurrence @param data_list: list to count items from ''' return dict((key, data_list.count(key)) for key in set(data_list))