From d55d81ca8241231b90f07592be1a8ab8af58a922 Mon Sep 17 00:00:00 2001
From: jezozwierzak <jezozwierzak@gmail.com>
Date: Thu, 23 Aug 2012 09:53:35 +0200
Subject: [PATCH] Changed CSVWriter to CSVTable
---
utils/CSVColumn.py | 166 ++++++++++++++++++++++++++++++++++++++
utils/CSVTable.py | 178 ++++++++++++++++++++++++++++++++++++++++
utils/CSVWriter.py | 192 --------------------------------------------
utils/chunk_eval.py | 95 ++++++++++++++--------
4 files changed, 404 insertions(+), 227 deletions(-)
create mode 100755 utils/CSVColumn.py
create mode 100755 utils/CSVTable.py
delete mode 100644 utils/CSVWriter.py
diff --git a/utils/CSVColumn.py b/utils/CSVColumn.py
new file mode 100755
index 0000000..ad551de
--- /dev/null
+++ b/utils/CSVColumn.py
@@ -0,0 +1,166 @@
+#!/usr/bin/python
+#-*- coding: utf-8 -*-
+'''
+Created on 09-08-2012
+
+@author: Adam Pawlaczek
+'''
+# Copyright (C) 2012 Adam Pawlaczek.
+# This program is free software; you can redistribute and/or modify it
+# under the terms of the GNU Lesser General Public License as published by the Free
+# Software Foundation; either version 3 of the License, or (at your option)
+# any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+# or FITNESS FOR A PARTICULAR PURPOSE.
+#
+# See the LICENCE and COPYING files for more details
+
+class CSVColumn:
+
+ def __init__(self, parent, name, separator, type):
+ self.parent = parent
+ self.name = name
+ self.type = type
+ self.width = len(name)
+ self.content = []
+ self.separator = separator
+
+ def addSubColumn(self, name, type = ''):
+ assert len(self.content) == 0 or self.hasSubColumns()
+ self.type = 'dict'
+ self.content.append(CSVColumn(self, name, self.separator, type))
+ self.recountWidths()
+
+ def insertValue(self, row, data, subColumn =''):
+
+ if self.hasSubColumns() and type(data).__name__ == 'dict':
+ for subColumn in self.content:
+ subColumn.insertValue(data[subColumn.name], row)
+ elif self.hasSubColumns():
+ for sub in self.content:
+ if sub.name == subColumn:
+ sub.content[row] = data
+ if type(data).__name__ == "float":
+ data_str = "%.4f"%data
+ else:
+ data_str = str(data)
+ if len(data_str) > sub.width:
+ sub.width = len(data_str)
+ break
+ elif subColumn == '':
+ self.content[row] = data
+ if type(data).__name__ == "float":
+ data_str = "%.4f"%data
+ else:
+ data_str = str(data)
+ if len(data_str) > self.width:
+ self.width = len(data_str)
+
+ def increment(self, row, subColumn=''):
+ if subColumn != '' and self.hasSubColumns():
+ for sub in self.content:
+ if sub.name == subColumn:
+ sub.content[row] += 1
+ if sub.type == "float":
+ data_str = "%.4f"%sub.content[row]
+ else:
+ data_str = str(sub.content[row])
+ if len(data_str) > sub.width:
+ self.width += len(data_str) - sub.width
+ sub.width = len(data_str)
+ break
+ else:
+ self.content[row] += 1
+
+
+ def getValue(self, row, subColumn =''):
+ if subColumn != '' and self.hasSubColumns():
+ for sub in self.content:
+ if sub.name == subColumn:
+ return sub.content[row]
+ else:
+ return self.content[row]
+
+ def addValue(self, data):
+ if self.hasSubColumns():
+ str_data = ''
+ for sub in self.content:
+ sub.addValue(data[sub.name])
+ str_data += str(data[sub.name]) + self.separator
+ if len(str_data) > self.width:
+ self.recountWidths()
+ else:
+ assert type(data).__name__ != 'dict', 'You added dict to column with no subcolumns'
+ if len(self.content) == 0:
+ self.type = type(data).__name__
+ self.content.append(data)
+ if len(str(data)) > self.width and self.isSubColumn():
+ self.parent.recountWidths()
+ elif len(str(data) + self.separator) > self.width:
+ self.width = len(str(data))
+
+ def countSum(self):
+ if self.hasSubColumns():
+ result = {}
+ for sub in self.content:
+ result[sub.name] = sub.countSum()
+ else:
+ if self.type != 'string':
+ result = 0
+ for row in self.content:
+ result += row
+ return result
+
+ def recountWidths(self):
+ subColsStr = ''
+ for subColumn in self.content:
+ subColsStr += subColumn.name + self.separator
+
+ if len(subColsStr) > self.width:
+ self.width = len(subColsStr)
+ for subColumn in self.content:
+ subColumn.width = len(subColumn.name + self.separator)
+ elif self.parent.rows > 0:
+ widths = 0
+ for i in range(0,len(self.content)): #Dla każdej podkolumny
+ for j in range(0, self.parent.rows): #Dla każdego wiersza
+ if i == len(self.content) - 1:
+ if self.width - widths > 0:
+ self.content[i].width = self.width - widths
+ else:
+ self.content[i].width = len(self.parent.ptr(self.content[i].content[j], 1))
+ break
+ elif len(self.parent.ptr(self.content[i].content[j], 1)) > self.content[i].width:
+ self.content[i].width = len(self.parent.ptr(self.content[i].content[j], 1))
+ widths += self.content[i].width
+ if widths > self.width:
+ self.width = widths
+
+ def hasSubColumns(self):
+ if len(self.content) > 0 and type(self.content[0]).__name__ == "instance":
+ return self.content[0].__class__ == CSVColumn
+ else:
+ return False
+
+ def isSubColumn(self):
+ if self.parent != '' and type(self.parent).__name__ == "instance" :
+ return self.parent.__class__ == CSVColumn
+ else:
+ return False
+
+ def fillZeros(self, rows):
+ if self.hasSubColumns():
+ for subcolumn in self.content:
+ subcolumn.fillZeros(rows)
+ else:
+ for i in range(0, rows):
+ if self.type == "float":
+ self.content.append(0.0)
+ elif self.type == "int":
+ self.content.append(0)
+ else:
+ self.content.append('')
+
+
\ No newline at end of file
diff --git a/utils/CSVTable.py b/utils/CSVTable.py
new file mode 100755
index 0000000..5cab654
--- /dev/null
+++ b/utils/CSVTable.py
@@ -0,0 +1,178 @@
+#!/usr/bin/python
+#-*- coding: utf-8 -*-
+'''
+Created on 09-08-2012
+
+@author: Adam Pawlaczek
+
+TODO: Ew. dodać float do obliczania AVG zamiast rzutowania na typ, w którym jest kolumna.
+
+'''
+# Copyright (C) 2012 Adam Pawlaczek.
+# This program is free software; you can redistribute and/or modify it
+# under the terms of the GNU Lesser General Public License as published by the Free
+# Software Foundation; either version 3 of the License, or (at your option)
+# any later version.
+#
+# This program is distributed in the hope that it will be useful, but
+# WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+# or FITNESS FOR A PARTICULAR PURPOSE.
+#
+# See the LICENCE and COPYING files for more details
+
+from CSVColumn import CSVColumn
+from operator import itemgetter
+
+class CSVTable:
+
+ def __init__(self, separator = ';'):
+ self.widths = []
+ self.content = []
+ self.rows = 0
+ self.columns = 0
+ self.separator = separator
+
+ def addColumn(self, name, type = ''):
+ for column in self.content:
+ assert column.name != name, 'Column with name: '+ name+ ' already exists'
+ column = CSVColumn(self, name, self.separator, type)
+ self.content.append(column)
+
+ if self.rows > 0 and type != 'dict':
+ column.fillZeros(self.rows)
+ self.columns += 1
+ return True
+
+ def addSubColumn(self, parentName, name, type = ''):
+ for column in self.content:
+ if column.name == parentName:
+ column.addSubColumn(name, type)
+ if len(column.content) > 1:
+ self.columns+=1
+ if self.rows > 0:
+ column.content[-1].fillZeros(self.rows)
+ return True
+ return False
+
+ def addRow(self, row={}):
+ for column in self.content:
+ if column.name != 'Nr' or 'Nr' in row.keys():
+ column.addValue(row[column.name])
+ else:
+ if self.rows > 0:
+ column.addValue(column.content[self.rows-1] + 1)
+ else:
+ column.addValue(1)
+ self.rows += 1
+
+ def addEmptyRow(self):
+ for column in self.content:
+ if column.name == "Nr":
+ if self.rows > 0:
+ column.addValue(column.content[self.rows-1] + 1)
+ else:
+ column.addValue(1)
+ else:
+ column.fillZeros(1)
+ self.rows += 1
+
+ def insertInColumn(self, columnName, row, data, subColumn= ''):
+ assert row < self.rows
+ for column in self.content:
+ if column.name == columnName:
+ column.insertValue(row, data, subColumn)
+
+ def increment(self, columnName, row, subColumn=''):
+ for column in self.content:
+ if column.name == columnName:
+ column.increment(row, subColumn)
+
+ def getValue(self, columnName, row, subColumn= ''):
+ assert row < self.rows
+ for column in self.content:
+ if column.name == columnName:
+ return column.getValue(row, subColumn)
+
+ def hasSubColumns(self):
+ for column in self.content:
+ if column.hasSubColumns():
+ return True
+ return False
+
+ def hasNrColumn(self):
+ return self.hasColumn("Nr")
+
+ def hasColumn(self, name):
+ for column in self.content:
+ if column.name == name:
+ return True
+ return False
+
+ def countAvg(self):
+ results = {}
+ for column in self.content:
+ results[column.name] = column.countSum()
+ if column.type == 'int' or column.type == 'float':
+ results[column.name] = results[column.name] / self.rows
+ elif column.type == 'dict':
+ for subColumn in results[column.name].keys():
+ results[column.name][subColumn] = results[column.name][subColumn] / self.rows
+ if self.hasNrColumn():
+ results['Nr'] = 'AVG'
+ self.addRow(results)
+
+
+ def countSum(self):
+ results = {}
+ for column in self.content:
+ results[column.name] = column.countSum()
+ if self.hasNrColumn():
+ results['Nr'] = 'SUM'
+ self.addRow(results)
+
+ ''' Helping functions '''
+
+ def __repeat_to_length(self, string_to_expand, length):
+ return (string_to_expand * ((length/len(string_to_expand))+1))[:length]
+
+ def ptr(self, data, width):
+ if type(data).__name__=='int':
+ return '{0:{width}{base}}'.format(data, base='d', width=width) + self.separator
+ elif type(data).__name__=='float':
+ return '{0:{width}{base}}'.format(data, base='.4f', width=width) + self.separator
+ else:
+ return '{0:{width}{base}}'.format(data, base='s', width=width) + self.separator
+
+
+ def __str__(self):
+ result = ''
+ #HEADERS
+ for column in self.content:
+ result += self.ptr(column.name, column.width)
+ if column.hasSubColumns():
+ for i in range(1,len(column.content)):
+ result += self.separator
+ result += '\n'
+ #SUBHEADERS
+ for column in self.content:
+ if self.hasSubColumns():
+ if column.hasSubColumns():
+ for subColumn in column.content:
+ if len(column.content) > 1:
+ result += self.ptr(subColumn.name, subColumn.width)
+ else:
+ result += self.ptr(subColumn.name, column.width)
+ else:
+ result += self.ptr(' ', column.width)
+ result += '\n'
+ #DATA
+ for i in range(0, self.rows): #Rows iteration
+ for column in self.content: #Column iteration
+ if column.hasSubColumns():
+ for subColumn in column.content:
+ result += self.ptr(subColumn.content[i], subColumn.width)
+ else:
+ result += self.ptr(column.content[i], column.width)
+ result += '\n'
+ return result
+
\ No newline at end of file
diff --git a/utils/CSVWriter.py b/utils/CSVWriter.py
deleted file mode 100644
index deb2e20..0000000
--- a/utils/CSVWriter.py
+++ /dev/null
@@ -1,192 +0,0 @@
-#!/usr/bin/python
-#-*- coding: utf-8 -*-
-'''
-Created on 03-08-2012
-
-@author: jezozwierzak
-'''
-
-class CSVWriter:
-
- def __init__(self, separator = ';'):
- self.widths = []
- self.list = []
- self.rows = 0
- self.columns = 0
- self.separator = separator
-
- def addSubColumn(self, parentIndex, name):
- parentColumn = self.list[parentIndex];
- parentColumn.append([name])
-
- subColsStr = ''
- for i in range(1,len(parentColumn)):
- subColsStr += parentColumn[i][0] + self.separator
-
- if len(subColsStr) > self.widths[parentIndex]:
- self.widths[parentIndex] = len(subColsStr) - 1
-
- if len(parentColumn[1:]) > 1:
- self.columns+=1
-
- def addSubColumnByName(self, parentName, name):
- assert self.rows == 0, 'You have to add all Column names before adding rows'
- parentIndex = self.columnIndex(parentName)
- self.addSubColumn(parentIndex, name)
-
- def addSubColumnsByName(self, parentName, names=[]):
- for name in names:
- self.addSubColumnByName(parentName, name)
-
- def addSubColumns(self, parentIndex, names=[]):
- for name in names:
- self.addSubColumn(parentIndex, name)
-
- def addColumn(self, name):
- assert self.rows == 0, 'You have to add all Column names before adding rows'
- self.list.append([name])
- self.widths.append(len(name))
- self.columns+=1
-
- def addColumns(self, names=[]):
- for name in names:
- self.addColumn(name)
-
- def addRow(self, row=[]):
- assert len(row) == len(self.list), 'Wrong number of columns in row'
-
- for i in range(0,len(self.list)):
- column = self.list[i]
- if len(column) > 1 and type(column[1]).__name__ == 'list':
- #Adding data to subcolumns
- assert len(row[i]) == len(column) - 1, 'Wrong number of subColumns in column ' + column[0]
-
- for j in range(0,len(row[i])):
- column[j+1].append(row[i][j])
-
- subColsStr = ''
- for j in range(0,len(row[i])):
- subColsStr += '{0:{base}}'.format(row[i][j], base='.4f') + self.separator + ' '
-
- if len(subColsStr) > self.widths[i]:
- self.widths[i] = len(subColsStr) - 1
- else:
- #Adding data to column
- column.append(row[i])
-
- if len(str(row[i])) > self.widths[i]:
- self.widths[i] = len(str(row[i]))
-
- self.rows+=1
-
- def allWidth(self):
- sum = 0
- for width in self.widths:
- sum += width
- return width
-
- def columnIndex(self, name):
- for column in self.list:
- if column[0] == name:
- return self.list.index(column)
-
- def hasSubColumns(self):
- for column in self.list:
- if len(column) > 1 and type(column[1]).__name__ == 'list':
- return True
- return False
-
- def hasColumnSubColumns(self, index):
- column = self.list[index]
- return len(column) > 1 and type(column[1]).__name__ == 'list'
-
- def repeat_to_length(self, string_to_expand, length):
- return (string_to_expand * ((length/len(string_to_expand))+1))[:length]
-
- def count_avg(self, ):
- results = []
- if not self.hasSubColumns():
- for i in range(0,len(self.list)): #Iterowanie po kolumnach
- results.append(0)
- for j in range(1,1+self.rows): # Iterowanie po wierszach
- results[i]+= self.list[i][j]
- results[i]/=self.rows
- else:
- for i in range(0,len(self.list)): #Iterowanie po kolumnach
-
- if self.hasColumnSubColumns(i):
- subResults = []
- for k in range(0,len(self.list[i][1:])):
- subColumn = self.list[i][1:][k]
- subResults.append(0)
- for j in range(1,1+self.rows): # Iterowanie po wierszach
- subResults[k]+= subColumn[j]
- subResults[k]/=self.rows
- results.append(subResults)
- else:
- results.append(0)
- for j in range(1,1+self.rows): # Iterowanie po wierszach
- results[i]+= self.list[i][j]
- results[i]/=self.rows
- results = results[1:]
- results[:0] = ['AVG']
- self.addRow(results)
-
- def __str__(self):
- result = ''
- if not self.hasSubColumns():
- for j in range(0,1+self.rows): # Iterowanie po wierszach
- for i in range(0, len(self.list)): #Iterowanie po kolumnach
- if type(self.list[i][j]).__name__=='int':
- result += '{0:{width}{base}}'.format(self.list[i][j], base='d', width=self.widths[i]) + self.separator
- elif type(self.list[i][j]).__name__=='float':
- result += '{0:{width}{base}}'.format(self.list[i][j], base='.4f', width=self.widths[i]) + self.separator
- else:
- result += '{0:{width}{base}}'.format(self.list[i][j], base='s', width=self.widths[i]) + self.separator
- result += '\n'
- else:
- #Printing Thead
- thead = zip(*self.list)[0]
- for i in range(0,len(thead)):
- if self.hasColumnSubColumns(i):
- numberOfColumns = len(self.list[i][1:])
- result += '{0:{width}{base}}'.format(thead[i], base='s', width=self.widths[i]-numberOfColumns+1) + self.separator
- for j in range(1,numberOfColumns):
- result += self.separator
- else:
- result += '{0:{width}{base}}'.format(thead[i], base='s', width=self.widths[i]) + self.separator
- result += '\n'
- #Printing subTheads:
- for i in range(0,len(self.list)):
- if self.hasColumnSubColumns(i):
- numberOfColumns = len(self.list[i][1:])
- for subColumn in self.list[i][1:]:
- result += '{0:{width}{base}}'.format(subColumn[0], base='s', width=(self.widths[i]/numberOfColumns)) + self.separator
- else:
- result += '{0:{width}{base}}'.format('', base='s', width=self.widths[i]) + self.separator
- result += '\n'
- #Printing Data
-
- for j in range(1,1+self.rows): # Iterowanie po wierszach
- for i in range(0, len(self.list)): #Iterowanie po kolumnach
- if self.hasColumnSubColumns(i):
-
- for subcolumns in self.list[i][1:]:
- if type(subcolumns[j]).__name__=='int':
- result += '{0:{width}{base}}'.format(subcolumns[j], base='d', width=(self.widths[i]/numberOfColumns)) + self.separator
- elif type(subcolumns[j]).__name__=='float':
- result += '{0:{width}{base}}'.format(subcolumns[j], base='.4f', width=(self.widths[i]/numberOfColumns)) + self.separator
- else:
- result += '{0:{width}{base}}'.format(subcolumns[j], base='s', width=(self.widths[i]/numberOfColumns)) + self.separator
- else:
-
- if type(self.list[i][j]).__name__=='int':
- result += '{0:{width}{base}}'.format(self.list[i][j], base='d', width=self.widths[i]) + self.separator
- elif type(self.list[i][j]).__name__=='float':
- result += '{0:{width}{base}}'.format(self.list[i][j], base='.4f', width=self.widths[i]) + self.separator
- else:
- result += '{0:{width}{base}}'.format(self.list[i][j], base='s', width=self.widths[i]) + self.separator
- result += '\n'
- return result
-
-
diff --git a/utils/chunk_eval.py b/utils/chunk_eval.py
index 665d517..c0ac9e9 100755
--- a/utils/chunk_eval.py
+++ b/utils/chunk_eval.py
@@ -2,6 +2,8 @@
#-*- coding: utf-8 -*-
'''
Created on 01-08-2012
+
+@author: Adam Pawlaczek
'''
# Copyright (C) 2012 Adam Pawlaczek.
# This program is free software; you can redistribute and/or modify it
@@ -15,6 +17,12 @@ Created on 01-08-2012
#
# See the LICENCE and COPYING files for more details
+from optparse import OptionParser
+import corpus2
+import sys, os
+from CSVTable import CSVTable
+import codecs
+
descr = """%prog [options] CHUNKED REF
Reads the two chunk-annotated corpora: CHUNKED (chunker output) and REF
@@ -26,10 +34,6 @@ for the following settings:
NOTE: this script treats discontinuous chunks as whole annotations.
"""
-from optparse import OptionParser
-import corpus2
-import sys, os
-from CSVWriter import CSVWriter
class Stats:
def __init__(self):
@@ -58,13 +62,17 @@ class Stats:
self.head_hits += len(ch.intersection(ref))
def getPRF(self, hits):
+ result = {}
p = 0.0 if self.ch_chunks == 0 else 100.0 * hits / self.ch_chunks
r = 0.0 if self.ref_chunks == 0 else 100.0 * hits / self.ref_chunks
f = 0.0 if p + r == 0.0 else 2.0 * p * r / (p + r)
- return [p, r, f]
+ result['P'] = p
+ result['R'] = r
+ result['F'] = f
+ return result
def getStats(self):
- return [self.getPRF(self.chunk_hits)]
+ return self.getPRF(self.chunk_hits)
def get_annots(sent, chan_name):
# wrap the sentence as an AnnotatedSentence
@@ -106,12 +114,18 @@ def go():
ch_path, ref_path = args
main(ch_path, ref_path, options.chunk_names, options.input_format, options.out_path, options.tagset, options.verbose, options.folds)
-def main(ch_path, ref_path, chan_name, input_format, out_path, tagset, verbose, folds):
+def main(ch_path, ref_path, chan_names, input_format, out_path, tagset, verbose, folds):
+
+ chan_names = chan_names.split(",")
- csvWriter = CSVWriter(",")
-
- csvWriter.addColumns(["Nr ","Chunk"])
- csvWriter.addSubColumnsByName("Chunk", ["P", "R", "F"])
+ csvTable = CSVTable(";")
+ csvTable.addColumn('Nr')
+
+ for chan_name in chan_names:
+ csvTable.addColumn(chan_name)
+ csvTable.addSubColumn(chan_name, "P", type="float")
+ csvTable.addSubColumn(chan_name, "R", type="float")
+ csvTable.addSubColumn(chan_name, "F", type="float")
tagset = corpus2.get_named_tagset(tagset)
@@ -122,35 +136,46 @@ def main(ch_path, ref_path, chan_name, input_format, out_path, tagset, verbose,
else:
ch_path_fold = ch_path
ref_path_fold = ref_path
-
- ch_rdr = corpus2.TokenReader.create_path_reader(
+
+ results = {}
+
+ for chan_name in chan_names:
+
+ ch_rdr = corpus2.TokenReader.create_path_reader(
input_format, tagset, ch_path_fold)
- ref_rdr = corpus2.TokenReader.create_path_reader(
+ ref_rdr = corpus2.TokenReader.create_path_reader(
input_format, tagset, ref_path_fold)
-
- stats = Stats()
-
- while True:
- # iterate over paragraphs (note that they are called "chunks" here)
- ref_chunk = ref_rdr.get_next_chunk()
- ch_chunk = ch_rdr.get_next_chunk()
- assert (not ref_chunk) == (not ch_chunk), 'corpora of different length'
- if not ref_chunk:
- break # end of input
+ stats = Stats()
- # process each sentence separately
- for ch_sent, ref_sent in zip(ch_chunk.sentences(), ref_chunk.sentences()):
- assert ch_sent.size() == ref_sent.size()
- ch_annots = get_annots(ch_sent, chan_name)
- ref_annots = get_annots(ref_sent, chan_name)
- stats.update(ch_annots, ref_annots)
+ while True:
+ # iterate over paragraphs (note that they are called "chunks" here)
+ ref_chunk = ref_rdr.get_next_chunk()
+ ch_chunk = ch_rdr.get_next_chunk()
+ assert (not ref_chunk) == (not ch_chunk), 'corpora of different length'
+
+ if not ref_chunk:
+ break # end of input
+
+ # process each sentence separately
+ for ch_sent, ref_sent in zip(ch_chunk.sentences(), ref_chunk.sentences()):
+ assert ch_sent.size() == ref_sent.size()
+ ch_annots = get_annots(ch_sent, chan_name)
+ ref_annots = get_annots(ref_sent, chan_name)
+ stats.update(ch_annots, ref_annots)
- results = stats.getStats()
- results[:0] = [fold]
- csvWriter.addRow(results)
- csvWriter.count_avg()
- print csvWriter
+ results[chan_name] = stats.getStats()
+
+ csvTable.addRow(results)
+ if folds > 1:
+ csvTable.countAvg()
+
+ if out_path != '':
+ out = codecs.open(out_path, "w", "utf-8")
+ out.close()
+ else:
+ print csvTable
+
if __name__ == '__main__':
go()
\ No newline at end of file
--
GitLab