def convert_line(line):
new_line = []
# 0 1 2 3 4 5 6
# Dataset, Filename, #rows, Coder, %, Error Threshold, Window Param
#
new_line.append(line[CSVConstants.INDEX_DATASET])
new_line.append(line[CSVConstants.INDEX_FILENAME])
no_rows = line[CSVConstants.INDEX_NO_ROWS]
new_line.append(
MathUtils.str_to_int(no_rows) if isinstance(no_rows, int) else '')
new_line.append(line[CSVConstants.INDEX_ALGORITHM]) # Coder
threshold = line[CSVConstants.INDEX_THRESHOLD]
new_line.append(int(threshold) if len(threshold) > 0 else None) # %
new_line.append('') # Error Threshold
window = line[CSVConstants.INDEX_WINDOW]
new_line.append(
int(window) if len(window) > 0 else None) # Window Param
# 7 8 9 10 11 12
# Size (B), CR (%), Delta - Size (data), Delta - Size (mask), Delta - Size (total), Delta - CR (%), ...
#
for index in range(CSVConstants.INDEX_TOTAL_SIZE, len(line)):
if CSVConstants.is_percentage_index(index):
value = line[index]
if value.count('.') > 1: # e.g. "1.145.49"
value = value.replace('.', '',
1) # "1.145.49" => "1145.49"
new_line.append(float(value))
else:
new_line.append(MathUtils.str_to_int(line[index]))
return new_line
def generate_algorithm_line(self, name):
data = self.algorithms_data[name]
name = name.replace("Coder", "") # "CoderGAMPSLimit" => "CoderGAMPS"
name = name.replace("Limit", "") # "CoderGAMPS" => "GAMPS"
total = sum(data)
expected_total = 8*200 if name == "Total" else 200
# print(data)
# print(name)
# print(total)
# print(expected_total)
assert(total == expected_total)
line = [name]
total_percentage = 0
for value in data:
percentage = MathUtils.calculate_percentage(total, value, 1)
if name == "Total" and round(percentage,1) == 89.6:
percentage = 89.5
percentage = int(percentage) if int(percentage) == percentage else percentage
total_percentage += percentage
value_str = str(format(value,',d'))
if value_str != "0":
value_str += " (" + str(percentage) + "\%)"
line.append(value_str)
line = LatexUtils.format_line(line)
if name == "GAMPS": # last line before total
line += "\hline"
# print(name)
# print(round(total_percentage,1))
assert(round(total_percentage,1) == 100.0)
return line
def calculate_relative_diff(row_df_pca, row_df_apca, col_index):
data_column_key = ResultsToDataframe.data_column_key(col_index)
size_pca, size_apca = row_df_pca[data_column_key], row_df_apca[
data_column_key]
relative_diff = MathUtils.relative_difference(size_pca, size_apca)
coder_name = 'PCA' if size_pca < size_apca else 'APCA'
return round(relative_diff, 2), coder_name
def set_percentages(line, line_total):
assert (len(line) == len(line_total))
for index in range(len(line)):
if CSVConstants.is_percentage_index(index):
total, value = line_total[index - 1], line[index - 1]
percentage = MathUtils.calculate_percent(total, value)
line[index] = percentage
return line
def __gzip_new_line(gzip_results_reader, row):
dataset_name, filename = row
new_line = [dataset_name, filename, None, 'CoderGZIP'] + [None] * 9
for column_name in ExperimentsUtils.COLUMN_INDEXES[dataset_name]:
gzip_bits, base_bits = gzip_results_reader.gzip_and_base_bits(
dataset_name, filename, column_name)
percentage = MathUtils.calculate_percentage(
base_bits, gzip_bits, 5)
new_line += [gzip_bits, 0, gzip_bits, percentage]
return new_line
def generate_dataset_line(self, name):
data = self.datasets_data[name]
dataset_key = LatexUtils.get_dataset_key(name)
gaps_info = ExperimentsUtils.get_gaps_info(name)
assert (data['zero'] == 0)
total = data['negative'] + data['positive']
percentage = MathUtils.calculate_percentage(total, data['positive'], 2)
percentage = int(percentage) if int(
percentage) == percentage else round(percentage, 1)
outperform_str = str(data['positive']) + "/" + str(total) + " (" + str(
percentage) + "\%)"
range_str = self.range_str(data)
return LatexUtils.format_line(
[dataset_key, gaps_info, outperform_str, range_str])
def _calculate_relative_differences(self):
for row in self.data_rows:
row['relative_diffs'] = []
row['relative_diffs_string'] = []
for threshold_index in range(len(ExperimentsUtils.THRESHOLDS)):
for row in self.data_rows:
smallest = self._smallest_total_bits(threshold_index)
total_bits = row['total_bits'][threshold_index]
rd = MathUtils.relative_difference(total_bits, smallest)
rd_string = self.format_rd(rd)
row['relative_diffs'].append(rd)
row['relative_diffs_string'].append(rd_string)
self._calculate_worst_relative_diff()
def calculate_RD(row_df_best, row_df_compare, col_index):
data_column_key = ResultsToDataframe.data_column_key(col_index)
size_best, size_compare = row_df_best[data_column_key], row_df_compare[
data_column_key]
relative_diff = MathUtils.relative_difference(size_compare, size_best)
return round(relative_diff, 2)