def test_build_with_options(self):
doc = Document('With options', doc_type='standalone', options=['12pt', 'Spam'], egg=42)
assert doc.build(False, False, False) == cleandoc(r'''\documentclass[12pt, Spam, egg=42]{standalone}
\usepackage[utf8]{inputenc}
\usepackage[top=2.5cm, bottom=2.5cm, left=2.5cm, right=2.5cm]{geometry}
\begin{document}
\end{document}''')
def test_build_with_relative_path_from_source(self):
filepath = './some_doc_path/'
doc_name = 'Doc name'
doc = Document(doc_name, filepath=filepath)
doc += 'Some text'
try:
doc.build(show_pdf=False, build_from_dir='source')
assert os.path.exists(filepath + doc_name + '.tex')
assert os.path.exists(filepath + doc_name + '.pdf')
finally:
if os.path.exists(filepath): shutil.rmtree(filepath)
def test_deletes_files_all(self):
filepath = './some_doc_path/'
doc_name = 'doc_name'
doc = Document(doc_name, filepath=filepath)
doc += 'Some text'
try:
doc.build(show_pdf=False, delete_files='all')
assert not os.path.exists(filepath + doc_name + '.tex')
assert os.path.exists(filepath + doc_name + '.pdf')
finally:
shutil.rmtree('./some_doc_path/')
def test_build_to_other_relative_path(self):
filepath = './some_doc_path/'
doc_name = 'Doc name'
doc = Document(doc_name, filepath=filepath)
doc += 'Some text'
try:
doc.build(show_pdf=False)
assert os.path.exists(filepath + doc_name + '.tex')
assert os.path.exists(filepath + doc_name + '.pdf')
finally:
shutil.rmtree('./some_doc_path/')
def test_deletes_files_aux_and_log(self):
filepath = './some_doc_path/'
doc_name = 'doc_name'
doc = Document(doc_name, filepath=filepath)
doc += 'Some text'
try:
doc.build(show_pdf=False, delete_files=['aux', 'log'])
assert not os.path.exists(filepath + doc_name + '.aux')
assert not os.path.exists(filepath + doc_name + '.log')
assert os.path.exists(filepath + doc_name + '.pdf')
assert os.path.exists(filepath + doc_name + '.tex')
finally:
if os.path.exists(filepath): shutil.rmtree(filepath)
def test_preamble_appears_in_document(self):
doc = Document('test')
color = Color(1, 2, 3)
command = TexCommand('somecommand', 'param', options=[color])
doc += command
assert doc.build(False, False, False) == cleandoc(r'''
\documentclass{article}
\usepackage[utf8]{inputenc}
\usepackage[top=2.5cm, bottom=2.5cm, left=2.5cm, right=2.5cm]{geometry}
\definecolor{color1}{rgb}{1,2,3}
\begin{document}
\somecommand{param}[color1]
\end{document}''')
def test_preamble_appears_in_document(self):
doc = Document('test')
colored_text = textcolor(Color(1, 0, 0, color_name='my_color'), 'hello')
doc += colored_text
assert doc.build(False, False, False) == cleandoc(r'''
\documentclass{article}
\usepackage[utf8]{inputenc}
\usepackage[top=2.5cm, bottom=2.5cm, left=2.5cm, right=2.5cm]{geometry}
\usepackage[dvipsnames]{xcolor}
\definecolor{my_color}{rgb}{1,0,0}
\begin{document}
\textcolor{my_color}{hello}
\end{document}''')
def write(self, output_path):
if not os.path.exists(output_path):
os.makedirs(output_path)
doc = Document(filename='table4',
filepath=output_path,
doc_type='article',
options=('12pt', ))
doc.add_to_preamble(r"\usepgfplotslibrary{fillbetween}")
doc.add_to_preamble(r'\usepgfplotslibrary{colorbrewer}')
doc.add_to_preamble(r'\pgfplotsset{compat=1.15, colormap/Blues}')
sec = doc.new_section('All graphs')
self.write_CSLS(sec, output_path)
self.write_vocabulary_cutoff(sec, output_path)
self.write_stochastic(sec, output_path)
doc.build(save_to_disk=True, compile_to_pdf=False, show_pdf=False)
def plot_palette(
doc_name,
palette,
n_colors=5,
width=5,
height=1,
):
palette.n_colors = n_colors
palette.tex_colors = []
palette._init_colors()
color_width = width / n_colors
tikzpic = TexEnvironment('tikzpicture')
tikzpic.add_package('tikz')
for i, color in zip(range(n_colors), palette):
pos = i * color_width
tikzpic += f'\\fill[draw, {build(color, tikzpic)}] ({pos},0) rectangle ({pos+color_width},{height});'
doc = Document(doc_name, filepath='./palettes/', doc_type='standalone')
doc += tikzpic
doc.build(delete_files=['log', 'aux'])
def write(self, output_path):
if not os.path.exists(output_path):
os.makedirs(output_path)
doc = Document(filename='grid_search_experiments',
filepath=output_path,
doc_type='article',
options=('12pt', ))
doc.add_to_preamble(r"\usepgfplotslibrary{fillbetween}")
doc.add_to_preamble(r'\usepgfplotslibrary{colorbrewer}')
doc.add_to_preamble(
r'\pgfplotsset{compat=1.15, colormap/Blues, every axis/.append style={label style={font=\footnotesize}, tick label style={font=\footnotesize}}}'
)
sec = doc.new_section('All graphs')
self.write_CSLS(sec, output_path)
self.write_vocabulary_cutoff(sec, output_path)
self.write_stochastic(sec, output_path)
doc.build(save_to_disk=True, compile_to_pdf=False, show_pdf=False)
def test_build_with_body_and_packages(self):
doc = Document('With options', doc_type='standalone', options=['12pt', 'Spam'], egg=42)
doc.add_package('tikz')
sec = doc.new_section('Section', label='Section')
sec.add_text('Hey')
assert doc.build(False, False, False) == cleandoc(r'''\documentclass[12pt, Spam, egg=42]{standalone}
\usepackage[utf8]{inputenc}
\usepackage[top=2.5cm, bottom=2.5cm, left=2.5cm, right=2.5cm]{geometry}
\usepackage{tikz}
\begin{document}
\begin{section}{Section}
\label{section:Section}
Hey
\end{section}
\end{document}''')
from python2latex import Document, Plot
import numpy as np
# Create the document
filepath = './examples/simple matrix plot example'
filename = 'simple_matrix_plot_example'
doc = Document(filename,
doc_type='standalone',
filepath=filepath,
border='1cm')
# Create the data
X = np.linspace(-3, 3, 11)
Y = np.linspace(-3, 3, 21)
# Create a plot
plot = doc.new(
Plot(plot_name=filename,
plot_path=filepath,
as_float_env=False,
grid=False,
lines=False,
enlargelimits='false',
width=r'.5\textwidth',
height=r'.5\textwidth'))
XX, YY = np.meshgrid(X, Y)
Z = np.exp(
-(XX**2 + YY**2) / 6
).T # Transpose is necessary because numpy puts the x dimension along columns and y dimension along rows, which is the opposite of a standard graph.
plot.add_matrix_plot(X, Y, Z)
from colorspacious import cspace_converter
JCh2rgb = cspace_converter('JCh', 'sRGB1')
import numpy as np
from python2latex import Document, Plot, LinearColorMap, Palette
# Create the document
filepath = './examples/plot examples/custom colors and line labels example/'
filename = 'custom_colors_and_line_labels_example'
doc = Document(filename, doc_type='article', filepath=filepath)
# Create color map in JCh space, in which each parameter is linear with human perception
cmap = LinearColorMap(
color_anchors=[(20, 45, 135), (81, 99, 495)],
color_model='JCh',
color_transform=lambda color: np.clip(JCh2rgb(color), 0, 1))
# Create a dynamical palette which generates as many colors as needed from the cmap. Note that by default, the range of color used expands with the number of colors.
palette = Palette(colors=cmap, color_model='rgb')
pal = Palette(colors=cmap, n_colors=2)
# Create the data
X = np.linspace(-1, 1, 50)
Y = lambda c: np.exp(X * c) + c
# Let us compare the different color palettes generated for different number of line plots
for n_colors in [2, 3, 5, 10]:
# Create a plot
plot = doc.new(
Plot(
from python2latex import Document, Plot, Color
import numpy as np
# Create the document
filepath = './examples/more complex plot example/'
filename = 'more_complex_plot_example'
doc = Document(filename, doc_type='article', filepath=filepath)
sec = doc.new_section('More complex plot')
sec.add_text(
'This section shows how to make a more complex plot integrated directly into a tex file.'
)
# Create the data
X = np.linspace(0, 2 * np.pi, 100)
Y1 = np.sin(X)
Y2 = np.cos(X)
# Create a plot
plot = sec.new(Plot(plot_name=filename, plot_path=filepath))
plot.caption = 'More complex plot'
nice_blue = Color(.07, .22, .29, color_name='nice_blue')
nice_orange = Color(.85, .33, .28, color_name='nice_orange')
plot.add_plot(X, Y1, nice_blue, 'dashed',
legend='sine') # Add colors and legend to the plot
plot.add_plot(X, Y2, nice_orange, line_width='3pt', legend='cosine')
plot.legend_position = 'south east' # Place the legend where you want
# Add a label to each axis
plot.x_label = 'Radians'
from python2latex import Document
doc = Document(filename='simple_document_example', filepath='./examples/simple document example', doc_type='article', options=('12pt',))
doc.set_margins(top='3cm', bottom='3cm', margins='2cm')
sec = doc.new_section('Spam and Egg', label='spam_egg')
sec.add_text('The Monty Python slays the Spam and eats the Egg.')
tex = doc.build() # Builds to tex and compile to pdf
print(tex) # Prints the tex string that generated the pdf
# Second, obtain the data
n_datasets = 10
n_models = 4
n_draws = 5
data = np.full((n_datasets, n_models), 0, dtype=object)
for i, row in enumerate(data):
for j, _ in enumerate(row):
data[i, j] = MeanWithStd(np.random.rand(n_draws))
# Third, create the table
# Create a basic document
doc = Document(filename='mean_with_std_table_example',
filepath='examples/table examples',
doc_type='article',
options=('12pt', ))
# Create table
col, row = n_models + 1, n_datasets + 2
table = doc.new(Table(shape=(row, col), float_format='.3f'))
# Set a caption
table.caption = f'Mean test accuracy (and standard deviation) of {n_models} models on {n_datasets} datasets for {n_draws} different random states.'
table.caption_space = '10pt' # Space between table and caption.
# Set entries with slices
table[2:, 1:] = data
table[2:, 0] = [f'Dataset {i}' for i in range(n_datasets)]
table[0, 1:] = 'Models'
table[1, 1:] = [f'Model {i}' for i in range(n_models)]
from python2latex import Document, Table, italic
import numpy as np
doc = Document(filename='more_complex_table_from_numpy_array_example',
filepath='examples/table examples',
doc_type='article',
options=('12pt', ))
sec = doc.new_section('Testing tables from numpy array')
sec.add_text("This section tests tables from numpy array.")
col, row = 6, 4
data = np.random.rand(row, col)
table = sec.new(
Table(shape=(row + 2, col + 1), alignment='c', float_format='.2f'))
# Set a caption if desired
table.caption = 'Table from numpy array'
table.caption_space = '10pt' # Space between table and caption.
# Set entries with slices
table[2:, 1:] = data
# Overwrite data if needed, whatever the object type
table[2:, 1] = [i * 1000 for i in range(row)]
# Change format of cells easily
table[2:, 1].format_spec = '.0e' # Exponential format
# Apply custom functions on the cell content for flexibility
table[2, 1].apply_command(lambda value: f'${value}$')
def write(self, output_path):
if not os.path.exists(output_path):
os.makedirs(output_path)
doc = Document(filename='table2',
filepath=output_path,
doc_type='article',
options=('12pt', ))
sec = doc.new_section('Table 2')
col, row = 17, 17
table = sec.new(
LatexTable(shape=(row, col),
alignment=['l'] + ['c'] * 16,
float_format='.1f',
label='ablation_study'))
table.caption = self.CAPTION
table.label_pos = 'bottom'
# Main header
table[0, 1:5].multicell(bold('EN-DE'), h_align='c')
table[0, 5:9].multicell(bold('EN-ES'), h_align='c')
table[0, 9:13].multicell(bold('EN-FI'), h_align='c')
table[0, 13:17].multicell(bold('EN-IT'), h_align='c')
table[0, 1:5].add_rule(trim_left=True, trim_right='.3em')
table[0, 5:9].add_rule(trim_left='.3em', trim_right='.3em')
table[0, 9:13].add_rule(trim_left='.3em', trim_right='.3em')
table[0, 13:17].add_rule(trim_left='.3em', trim_right=True)
# Sub header
table[1, 1:17] = (['best', 'avg', 's', 't'] * 4)
table[1, 0:17].add_rule(trim_left=True, trim_right=True)
### Full system metrics
table[2, 0] = 'Full System'
table = self.write_original_row(table, 2,
self.ORIGINAL_RESULTS['Full System'])
experiment = self.experiments['Full System']
metrics = experiment.aggregate_runs()
table[3, 0] = bold('Reproduced')
table = self.write_new_row(table, 3, metrics)
table[3, 0:17].add_rule(trim_left=True, trim_right=True)
### Unsup. Init
table[4, 0] = '- Unsup. Init.'
table = self.write_original_row(table, 4,
self.ORIGINAL_RESULTS['Unsup. Init'])
experiment = self.experiments['Unsup. Init (Random)']
metrics = experiment.aggregate_runs()
table[5, 0] = bold('Rand.')
table = self.write_new_row(table, 5, metrics)
experiment = self.experiments['Unsup. Init (Random Cutoff)']
metrics = experiment.aggregate_runs()
table[6, 0] = bold('Rand. Cut.')
table = self.write_new_row(table, 6, metrics)
table[6, 0:17].add_rule(trim_left=True, trim_right=True)
### Stochastic
table[7, 0] = '- Stochastic'
table = self.write_original_row(table, 7,
self.ORIGINAL_RESULTS['Stochastic'])
experiment = self.experiments['Stochastic']
metrics = experiment.aggregate_runs()
table[8, 0] = bold('Reproduced')
table = self.write_new_row(table, 8, metrics)
table[8, 0:17].add_rule(trim_left=True, trim_right=True)
### Cutoff
table[9, 0] = '- Cutoff (k=100k)'
table = self.write_original_row(
table, 9, self.ORIGINAL_RESULTS['Cutoff (k=100k)'])
# experiment = self.experiments['Cutoff (k=100k)']
# metrics = experiment.aggregate_runs()
table[10, 0] = bold('Reproduced')
# table = self.write_new_row(table, 10, metrics)
table[10, 1:] = ['-'] * 16
table[10, 0:17].add_rule(trim_left=True, trim_right=True)
### CSLS
table[11, 0] = '- CSLS'
table = self.write_original_row(table, 11,
self.ORIGINAL_RESULTS['CSLS'])
experiment = self.experiments['CSLS']
metrics = experiment.aggregate_runs()
table[12, 0] = bold('Reproduced')
table = self.write_new_row(table, 12, metrics)
table[12, 0:17].add_rule(trim_left=True, trim_right=True)
### Bidrectional
table[13, 0] = '- Bidrectional'
table = self.write_original_row(table, 13,
self.ORIGINAL_RESULTS['Bidrectional'])
experiment = self.experiments['Bidrectional']
metrics = experiment.aggregate_runs()
table[14, 0] = bold('Reproduced')
table = self.write_new_row(table, 14, metrics)
table[14, 0:17].add_rule(trim_left=True, trim_right=True)
### Re-weighting
table[15, 0] = '- Re-weighting'
table = self.write_original_row(table, 15,
self.ORIGINAL_RESULTS['Re-weighting'])
experiment = self.experiments['Re-weighting']
metrics = experiment.aggregate_runs()
table[16, 0] = bold('Reproduced')
table = self.write_new_row(table, 16, metrics)
tex = doc.build(save_to_disk=True,
compile_to_pdf=False,
show_pdf=False)
def write(self, output_path):
if not os.path.exists(output_path):
os.makedirs(output_path)
experiment = self.experiments['Other Languages']
metrics = experiment.aggregate_runs()
random_experiment = self.experiments[
'Other Languages Unsup. Init (Random)']
random_metrics = random_experiment.aggregate_runs()
random_cutoff_experiment = self.experiments[
'Other Languages Unsup. Init (Random Cutoff)']
random_cutoff_metrics = random_cutoff_experiment.aggregate_runs()
stochastic_experiment = self.experiments['Other Languages Stochastic']
stochastic_metrics = stochastic_experiment.aggregate_runs()
csls_experiment = self.experiments['Other Languages CSLS']
csls_metrics = csls_experiment.aggregate_runs()
bidirectional_experiment = self.experiments[
'Other Languages Bidrectional']
bidirectional_metrics = bidirectional_experiment.aggregate_runs()
reweighting_experiment = self.experiments[
'Other Languages Re-weighting']
reweighting_metrics = reweighting_experiment.aggregate_runs()
doc = Document(filename='table3',
filepath=output_path,
doc_type='article',
options=('12pt', ))
sec = doc.new_section('Table 3')
col, row = 17, 9
table = sec.new(
LatexTable(shape=(row, col),
alignment=['l'] + ['c'] * 16,
float_format='.1f',
label='other_languages_results'))
table.caption = self.CAPTION
table.label_pos = 'bottom'
# Main header
table[0, 1:5].multicell(bold('EN-ET'), h_align='c')
table[0, 5:9].multicell(bold('EN-FA'), h_align='c')
table[0, 9:13].multicell(bold('EN-LV'), h_align='c')
table[0, 13:17].multicell(bold('EN-VI'), h_align='c')
table[0, 1:5].add_rule(trim_left=True, trim_right='.3em')
table[0, 5:9].add_rule(trim_left='.3em', trim_right='.3em')
table[0, 9:13].add_rule(trim_left='.3em', trim_right='.3em')
table[0, 13:17].add_rule(trim_left='.3em', trim_right=True)
# Sub header
table[1, 1:17] = (['best', 'avg', 's', 't'] * 4)
table[1, 0:17].add_rule(trim_left=True, trim_right=True)
table[2, 0] = bold('Vecmap')
table[2, 1] = np.max(metrics['accuracies']['et'])
table[2, 2] = np.average(metrics['accuracies']['et'])
table[2,
3] = np.sum(np.array(metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[2, 4] = np.average(metrics['times']['et'])
table[2, 5] = np.max(metrics['accuracies']['fa'])
table[2, 6] = np.average(metrics['accuracies']['fa'])
table[2,
7] = np.sum(np.array(metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[2, 8] = np.average(metrics['times']['fa'])
table[2, 9] = np.max(metrics['accuracies']['lv'])
table[2, 10] = np.average(metrics['accuracies']['lv'])
table[2,
11] = np.sum(np.array(metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[2, 12] = np.average(metrics['times']['lv'])
table[2, 13] = np.max(metrics['accuracies']['vi'])
table[2, 14] = np.average(metrics['accuracies']['vi'])
table[2,
15] = np.sum(np.array(metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[2, 16] = np.average(metrics['times']['vi'])
table[3, 0] = bold('- Unsupervised (Random)')
table[3, 1] = np.max(random_metrics['accuracies']['et'])
table[3, 2] = np.average(random_metrics['accuracies']['et'])
table[3, 3] = np.sum(
np.array(random_metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[3, 4] = np.average(random_metrics['times']['et'])
table[3, 5] = np.max(random_metrics['accuracies']['fa'])
table[3, 6] = np.average(random_metrics['accuracies']['fa'])
table[3, 7] = np.sum(
np.array(random_metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[3, 8] = np.average(random_metrics['times']['fa'])
table[3, 9] = np.max(random_metrics['accuracies']['lv'])
table[3, 10] = np.average(random_metrics['accuracies']['lv'])
table[3, 11] = np.sum(
np.array(random_metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[3, 12] = np.average(random_metrics['times']['lv'])
table[3, 13] = np.max(random_metrics['accuracies']['vi'])
table[3, 14] = np.average(random_metrics['accuracies']['vi'])
table[3, 15] = np.sum(
np.array(random_metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[3, 16] = np.average(random_metrics['times']['vi'])
table[4, 0] = bold('- Unsupervised (Random Cutoff)')
table[4, 1] = np.max(random_cutoff_metrics['accuracies']['et'])
table[4, 2] = np.average(random_cutoff_metrics['accuracies']['et'])
table[4, 3] = np.sum(
np.array(random_cutoff_metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[4, 4] = np.average(random_cutoff_metrics['times']['et'])
table[4, 5] = np.max(random_cutoff_metrics['accuracies']['fa'])
table[4, 6] = np.average(random_cutoff_metrics['accuracies']['fa'])
table[4, 7] = np.sum(
np.array(random_cutoff_metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[4, 8] = np.average(random_cutoff_metrics['times']['fa'])
table[4, 9] = np.max(random_cutoff_metrics['accuracies']['lv'])
table[4, 10] = np.average(random_cutoff_metrics['accuracies']['lv'])
table[4, 11] = np.sum(
np.array(random_cutoff_metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[4, 12] = np.average(random_cutoff_metrics['times']['lv'])
table[4, 13] = np.max(random_cutoff_metrics['accuracies']['vi'])
table[4, 14] = np.average(random_cutoff_metrics['accuracies']['vi'])
table[4, 15] = np.sum(
np.array(random_cutoff_metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[4, 16] = np.average(random_cutoff_metrics['times']['vi'])
table[5, 0] = bold('- Stochastic')
table[5, 1] = np.max(stochastic_metrics['accuracies']['et'])
table[5, 2] = np.average(stochastic_metrics['accuracies']['et'])
table[5, 3] = np.sum(
np.array(stochastic_metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[5, 4] = np.average(stochastic_metrics['times']['et'])
table[5, 5] = np.max(stochastic_metrics['accuracies']['fa'])
table[5, 6] = np.average(stochastic_metrics['accuracies']['fa'])
table[5, 7] = np.sum(
np.array(stochastic_metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[5, 8] = np.average(stochastic_metrics['times']['fa'])
table[5, 9] = np.max(stochastic_metrics['accuracies']['lv'])
table[5, 10] = np.average(stochastic_metrics['accuracies']['lv'])
table[5, 11] = np.sum(
np.array(stochastic_metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[5, 12] = np.average(stochastic_metrics['times']['lv'])
table[5, 13] = np.max(stochastic_metrics['accuracies']['vi'])
table[5, 14] = np.average(stochastic_metrics['accuracies']['vi'])
table[5, 15] = np.sum(
np.array(stochastic_metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[5, 16] = np.average(stochastic_metrics['times']['vi'])
table[6, 0] = bold('- CSLS')
table[6, 1] = np.max(csls_metrics['accuracies']['et'])
table[6, 2] = np.average(csls_metrics['accuracies']['et'])
table[6, 3] = np.sum(
np.array(csls_metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[6, 4] = np.average(csls_metrics['times']['et'])
table[6, 5] = np.max(csls_metrics['accuracies']['fa'])
table[6, 6] = np.average(csls_metrics['accuracies']['fa'])
table[6, 7] = np.sum(
np.array(csls_metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[6, 8] = np.average(csls_metrics['times']['fa'])
table[6, 9] = np.max(csls_metrics['accuracies']['lv'])
table[6, 10] = np.average(csls_metrics['accuracies']['lv'])
table[6, 11] = np.sum(
np.array(csls_metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[6, 12] = np.average(csls_metrics['times']['lv'])
table[6, 13] = np.max(csls_metrics['accuracies']['vi'])
table[6, 14] = np.average(csls_metrics['accuracies']['vi'])
table[6, 15] = np.sum(
np.array(csls_metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[6, 16] = np.average(csls_metrics['times']['vi'])
table[7, 0] = bold('- Bidirectional')
table[7, 1] = np.max(bidirectional_metrics['accuracies']['et'])
table[7, 2] = np.average(bidirectional_metrics['accuracies']['et'])
table[7, 3] = np.sum(
np.array(bidirectional_metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[7, 4] = np.average(bidirectional_metrics['times']['et'])
table[7, 5] = np.max(bidirectional_metrics['accuracies']['fa'])
table[7, 6] = np.average(bidirectional_metrics['accuracies']['fa'])
table[7, 7] = np.sum(
np.array(bidirectional_metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[7, 8] = np.average(bidirectional_metrics['times']['fa'])
table[7, 9] = np.max(bidirectional_metrics['accuracies']['lv'])
table[7, 10] = np.average(bidirectional_metrics['accuracies']['lv'])
table[7, 11] = np.sum(
np.array(bidirectional_metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[7, 12] = np.average(bidirectional_metrics['times']['lv'])
table[7, 13] = np.max(bidirectional_metrics['accuracies']['vi'])
table[7, 14] = np.average(bidirectional_metrics['accuracies']['vi'])
table[7, 15] = np.sum(
np.array(bidirectional_metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[7, 16] = np.average(bidirectional_metrics['times']['vi'])
table[8, 0] = bold('- Reweighting')
table[8, 1] = np.max(reweighting_metrics['accuracies']['et'])
table[8, 2] = np.average(reweighting_metrics['accuracies']['et'])
table[8, 3] = np.sum(
np.array(reweighting_metrics['accuracies']['et']) > 1.0) / len(
metrics['accuracies']['et'])
table[8, 4] = np.average(reweighting_metrics['times']['et'])
table[8, 5] = np.max(reweighting_metrics['accuracies']['fa'])
table[8, 6] = np.average(reweighting_metrics['accuracies']['fa'])
table[8, 7] = np.sum(
np.array(reweighting_metrics['accuracies']['fa']) > 1.0) / len(
metrics['accuracies']['fa'])
table[8, 8] = np.average(reweighting_metrics['times']['fa'])
table[8, 9] = np.max(reweighting_metrics['accuracies']['lv'])
table[8, 10] = np.average(reweighting_metrics['accuracies']['lv'])
table[8, 11] = np.sum(
np.array(reweighting_metrics['accuracies']['lv']) > 1.0) / len(
metrics['accuracies']['lv'])
table[8, 12] = np.average(reweighting_metrics['times']['lv'])
table[8, 13] = np.max(reweighting_metrics['accuracies']['vi'])
table[8, 14] = np.average(reweighting_metrics['accuracies']['vi'])
table[8, 15] = np.sum(
np.array(reweighting_metrics['accuracies']['vi']) > 1.0) / len(
metrics['accuracies']['vi'])
table[8, 16] = np.average(reweighting_metrics['times']['vi'])
tex = doc.build(save_to_disk=True,
compile_to_pdf=False,
show_pdf=False)
# Show the generated colors in squares using TikZ.
for n_colors in [2, 3, 4, 5, 6, 9]:
doc += f'{n_colors} colors: \\hspace{{10pt}}'
tikzpicture = doc.new(
TexEnvironment('tikzpicture', options=['baseline=-.5ex']))
for i, color in zip(range(n_colors), palette):
tikzpicture += Node((i, 0), fill=color)
doc += ' '
if __name__ == "__main__":
# Create document
filepath = './examples/plot examples/predefined palettes comparison/'
filename = 'PREDEFINED_PALETTES_comparison'
doc = Document(filename, doc_type='article', filepath=filepath)
# Insert title
center = doc.new(TexEnvironment('center'))
center += r"\huge \bf Predefined color maps and palettes"
doc += """\\noindent
python2latex provides three color maps natively. They are defined in the JCh axes of the CIECAM02 color model, which is linear to human perception of colors. Moreover, three ``dynamic'' palettes have been defined, one for each color map. They are dynamic in that the range of colors used to produce the palette changes with the number of colors needed. This allows for a good repartition of hues and brightness for all choices of number of colors.
All three color maps have been designed to be colorblind friendly for all types of colorblindness. To do so, all color maps are only increasing or decreasing in lightness, which helps to distinguish hues that may look similar to a colorblind. This also has the advantage that the palettes are also viable in levels of gray.
"""
# First section
sec = doc.new_section(r'The \texttt{holi} color map')
sec += """
The ``holi'' color map was designed to provide a set of easily distinguishable hues for any needed number of colors. It is optimized for palettes of 5 or 6 colors, but other numbers of color also generate very good palettes. It is colorblind friendly for all types of colorblindness for up to 5 colors, but can still be acceptable for more colors. The name ``holi'' comes from the Hindu festival of colors. This is the default color map of python2latex.
from python2latex import Document, Table
import numpy as np
# Create the document of type standalone
doc = Document(filename='simple_table_from_numpy_array_example', filepath='examples/table examples', doc_type='standalone', border='10pt')
# Create the data
col, row = 4, 4
data = np.random.rand(row, col)
# Create the table and add it to the document at the same time
table = doc.new(Table(shape=(row+2, col+1), as_float_env=False)) # No float environment in standalone documents
# Set entries with a slice directly from a numpy array!
table[2:,1:] = data
# Set a columns title as a multicell with a simple slice assignment
table[0,1:] = 'Col title'
# Set whole lines or columns in a single line with lists
table[1,1:] = [f'Col{i+1}' for i in range(col)]
table[2:,0] = [f'Row{i+1}' for i in range(row)]
# Add rules where you want
table[1,1:].add_rule()
# Automatically highlight the best value(s) inside the specified slice, ignoring text
for r in range(2,row+2):
table[r].highlight_best('high', 'bold') # Best per row
tex = doc.build()
print(tex)
from python2latex import Document, Plot import numpy as np # Document type 'standalone' will only show the plot, but does not support all tex environments. filepath = './examples/simple plot example/' filename = 'simple_plot_example' doc = Document(filename, doc_type='standalone', filepath=filepath) # Create the data X = np.linspace(0,2*np.pi,100) Y1 = np.sin(X) Y2 = np.cos(X) # Create a plot plot = doc.new(Plot(X, Y1, X, Y2, plot_path=filepath, as_float_env=False)) tex = doc.build()
from python2latex import Document, TexEnvironment
doc = Document(filename='unsupported_env_example',
doc_type='article',
filepath='examples/unsupported env example',
options=('12pt', ))
sec = doc.new_section('Unsupported env')
sec.add_text("This section shows how to create unsupported env if needed.")
sec.add_package(
'amsmath') # Add needed packages in any TexEnvironment, at any level
align = sec.new(TexEnvironment('align', label='align_label'))
align.add_text(r"""e^{i\pi} &= \cos \pi + i \sin \pi\\
&= -1""") # Use raw strings to alleviate tex writing
tex = doc.build()
print(tex)
from python2latex import Document, Section, Subsection, TexEnvironment
doc = Document(filename='binding_objects_to_environments_example', filepath='./examples/binding objects to environments example', doc_type='article', options=('12pt',))
section = doc.bind(Section) # section is now a new class that creates Section instances that are automatically appended to 'doc'
sec1 = section('Section 1', label='sec1') # Equivalent to: sec1 = doc.new(Section('Section 1', label='sec1'))
sec1.add_text("All sections created with ``section'' will be automatically appended to the document body!")
subsection, texEnv = sec1.bind(Subsection, TexEnvironment) # 'bind' supports multiple classes in the same call
eq1 = texEnv('equation')
eq1.add_text(r'e^{i\pi} = -1')
eq2 = texEnv('equation')
eq2 += r'\sum_{n=1}^{\infty} n = -\frac{1}{12}' # The += operator calls is the same as 'add_text'
sub1 = subsection('Subsection 1 of section 1')
sub1 += 'Text of subsection 1 of section 1.'
sec2 = section('Section 2', label='sec2')
sec2 += "sec2 is also appended to the document after sec1."
tex = doc.build() # Builds to tex and compile to pdf
print(tex) # Prints the tex string that generated the pdf
enlargelimits='false',
y_dir='reverse',
axis_x='top',
x_tick_label_style='{font=\\footnotesize}',
y_tick_label_style='{font=\\footnotesize}',
colorbar_style=
'{ylabel=Attention, ylabel style={rotate=180}, yticklabels={0.00,0.25,0.50,0.75,1.00}, ytick={0,0.25,0.5,0.75,1}}'
)
plot.add_matrix_plot(X, Y, Z)
for i, zz in enumerate(Z):
for j, z in enumerate(zz):
if z > .6:
plot.axis += f'\\node[white] at ({i},{j}) {{\\scriptsize {z:.2f} }};'
plot.add_to_preamble(cmap_def)
plot.x_ticks = range(len(sentences[0]))
plot.x_ticks_labels = sentences[0]
plot.y_ticks = range(len(sentences[1]))
plot.y_ticks_labels = sentences[1]
doc = Document('heatmap_example',
filepath='./examples/',
doc_type='standalone')
doc.add_package('xcolor')
doc += plot
doc.build(delete_files=['log', 'aux'])
def write(self, output_path):
if not os.path.exists(output_path):
os.makedirs(output_path)
experiment = self.experiments['Reproduced Results']
metrics = experiment.aggregate_runs()
doc = Document(filename='table1',
filepath=output_path,
doc_type='article',
options=('12pt', ))
sec = doc.new_section('Table 1')
col, row = 17, 4
table = sec.new(
LatexTable(shape=(row, col),
alignment=['l'] + ['c'] * 16,
float_format='.1f',
label='original_results'))
table.caption = self.CAPTION
table.label_pos = 'bottom'
# Main header
table[0, 1:5].multicell(bold('EN-DE'), h_align='c')
table[0, 5:9].multicell(bold('EN-ES'), h_align='c')
table[0, 9:13].multicell(bold('EN-FI'), h_align='c')
table[0, 13:17].multicell(bold('EN-IT'), h_align='c')
table[0, 1:5].add_rule(trim_left=True, trim_right='.3em')
table[0, 5:9].add_rule(trim_left='.3em', trim_right='.3em')
table[0, 9:13].add_rule(trim_left='.3em', trim_right='.3em')
table[0, 13:17].add_rule(trim_left='.3em', trim_right=True)
# Sub header
table[1, 1:17] = (['best', 'avg', 's', 't'] * 4)
table[1, 0:17].add_rule(trim_left=True, trim_right=True)
table[2, 0] = 'Original'
table[2, 1] = self.ORIGINAL_RESULTS['de']['best']
table[2, 2] = self.ORIGINAL_RESULTS['de']['avg']
table[2, 3] = self.ORIGINAL_RESULTS['de']['successful']
table[2, 4] = self.ORIGINAL_RESULTS['de']['time']
table[2, 5] = self.ORIGINAL_RESULTS['es']['best']
table[2, 6] = self.ORIGINAL_RESULTS['es']['avg']
table[2, 7] = self.ORIGINAL_RESULTS['es']['successful']
table[2, 8] = self.ORIGINAL_RESULTS['es']['time']
table[2, 9] = self.ORIGINAL_RESULTS['fi']['best']
table[2, 10] = self.ORIGINAL_RESULTS['fi']['avg']
table[2, 11] = self.ORIGINAL_RESULTS['fi']['successful']
table[2, 12] = self.ORIGINAL_RESULTS['fi']['time']
table[2, 13] = self.ORIGINAL_RESULTS['it']['best']
table[2, 14] = self.ORIGINAL_RESULTS['it']['avg']
table[2, 15] = self.ORIGINAL_RESULTS['it']['successful']
table[2, 16] = self.ORIGINAL_RESULTS['it']['time']
table[3, 0] = bold('Reproduced')
table[3, 1] = np.max(metrics['accuracies']['de'])
table[3, 2] = np.average(metrics['accuracies']['de'])
table[3,
3] = np.sum(np.array(metrics['accuracies']['de']) > 1.0) / len(
metrics['accuracies']['de'])
table[3, 4] = np.average(metrics['times']['de'])
table[3, 5] = np.max(metrics['accuracies']['es'])
table[3, 6] = np.average(metrics['accuracies']['es'])
table[3,
7] = np.sum(np.array(metrics['accuracies']['es']) > 1.0) / len(
metrics['accuracies']['es'])
table[3, 8] = np.average(metrics['times']['es'])
table[3, 9] = np.max(metrics['accuracies']['fi'])
table[3, 10] = np.average(metrics['accuracies']['fi'])
table[3,
11] = np.sum(np.array(metrics['accuracies']['fi']) > 1.0) / len(
metrics['accuracies']['fi'])
table[3, 12] = np.average(metrics['times']['fi'])
table[3, 13] = np.max(metrics['accuracies']['it'])
table[3, 14] = np.average(metrics['accuracies']['it'])
table[3,
15] = np.sum(np.array(metrics['accuracies']['it']) > 1.0) / len(
metrics['accuracies']['it'])
table[3, 16] = np.average(metrics['times']['it'])
tex = doc.build(save_to_disk=True,
compile_to_pdf=False,
show_pdf=False)
from python2latex import Document, Plot, Color
import numpy as np
# Create the document
filepath = './examples/plot examples/more complex matrix plot example'
filename = 'more_complex_matrix_plot_example'
doc = Document(filename, doc_type='article', filepath=filepath)
sec = doc.new_section('More complex matrix plot')
sec.add_text(
'This section shows how to make a more complex matrix plot integrated directly into a tex file.'
)
# Adding necessary library to preamble for colormaps
doc.add_to_preamble(r'\usepgfplotslibrary{colorbrewer}')
doc.add_to_preamble(r'\pgfplotsset{compat=1.15, colormap/Blues-3}')
# Create the data
X = np.array([0.05, 0.1, 0.2])
Y = np.array([1.5, 2.0, 3.0, 4.0])
Z = np.random.random((3, 4))
# Create a plot
plot = sec.new(
Plot(plot_name=filename,
plot_path=filepath,
grid=False,
lines=False,
enlargelimits='false',
width=r'.6\textwidth',
height=r'.8\textwidth'))
plot.caption = 'Matrix plot of some random numbers as probabilities'
def default_doc():
return Document('Default')
xticklabel_style='{font=\\footnotesize}',
xtick_style='{draw=none}',
ytick_style='{draw=none}',
)
plot.axis.options += (
'nodes near coords',
'every node near coord/.append style={font=\\scriptsize}',
)
plot.add_plot(x, y1, 'fill', draw='none', legend='Modèle 1')
plot.add_plot(x, y2, 'fill', draw='none', legend='Modèle 2')
plot.legend_position = 'north west'
plot.y_min = 0
plot.y_max = 79
plot.x_min = 0.5
plot.x_ticks = x
plot.x_ticks_labels = labels
plot.x_label = 'Datasets'
plot.y_label = "Précision (\%)"
doc = Document('bar_chart_example',
filepath='./examples/',
doc_type='standalone')
doc += plot
doc.build(delete_files=['log', 'aux'])
from matplotlib.colors import hsv_to_rgb, rgb_to_hsv
"""
In this example, we explore color maps and palettes.
A color map is understood as a function taking as input a scalar between 0 and 1 and outputs and color in some format. The class LinearColorMap takes as input a sequence of colors and interpolates linearly the colors in-between to yield a color map.
A palette is simply a collection of colors. python2latex defines a Palette object that handles the boilerplating related to the creation of Color objects from a sequence of colors or from a color map. The Palette object modifies dynamically the colors generated according to the number of colors needed so that colors never repeat, as opposed to a standard palette which loops back to the beginning when colors are exhausted.
"""
# First create a conversion function
JCh2rgb = lambda color: np.clip(cspace_convert(color, 'JCh', 'sRGB1'), 0, 1)
rgb2JCh = cspace_converter('sRGB1', 'JCh')
# Create document
filepath = './examples/plot examples/JCh vs hsb color space/'
filename = 'JCh_vs_hsb_color_space'
doc = Document(filename, doc_type='article', filepath=filepath)
# Let us create a color map in the JCh color model, which parametrizes the colors according to human perception of colors instead of actual physical properties of light.
# Choose the color anchors of the color map defined in the JCh color space
color1_hsb = [.31, .9, .3]
color1_JCh = rgb2JCh(hsv_to_rgb(color1_hsb))
color2_hsb = [.31, .9, 1] # Same color with different brightness
color2_JCh = rgb2JCh(hsv_to_rgb(color2_hsb))
# Add full hue circle for color interpolation
color2_hsb[0] += 1
color2_JCh[2] += 360
# Create the color maps
cmap_JCh = LinearColorMap(color_anchors=[color1_JCh, color2_JCh],
color_model='JCh')
