def plot_centre_dist(self, thresh=2, show_threads=True, **kwargs):
"""Plots time elapsed since last comment for each participant"""
project, show, _ = ac.handle_kwargs(**kwargs)
data_high, data_low = self.__get_centre_distances(
thresh, split=True)
# set up and create plots
plt.style.use(SETTINGS['style'])
_, axes = plt.subplots()
colors_high = ac.color_list(
self.author_frame.loc[data_high.columns, 'color'],
SETTINGS['vmin'], SETTINGS['vmax'],
cmap=CMAP)
colors_low = ac.color_list(
self.author_frame.loc[data_low.columns, 'color'],
SETTINGS['vmin'], SETTINGS['vmax'],
cmap=CMAP)
data_high.plot(ax=axes, color=colors_high, legend=False)
data_low.plot(ax=axes, alpha=.1, color=colors_low, legend=False)
axes.set_ylabel("Days elapsed since last comment")
axes.set_title("Distance from centre of discussion\n{}".format(
project))
axes.xaxis.set_ticks_position('bottom')
axes.yaxis.set_ticks_position('left')
if show_threads:
self.__show_threads(axes)
ac.show_or_save(show)
def plot_activity_prop(self, **kwargs):
"""Shows plot of number of comments (bar) and proportion
level-1 / higher-level comment (line) for all authors"""
project, show, fontsize = ac.handle_kwargs(**kwargs)
plt.style.use(SETTINGS['style'])
cols = self.author_frame.columns[
self.author_frame.columns.str.startswith('level')].tolist()
data = self.author_frame[cols].copy()
data['proportion'] = (data[cols[1:]].sum(axis=1) /
data[cols].sum(axis=1))
colors = [plt.cm.Set1(20 * i) for i in range(len(data.index))]
axes = data[cols].plot(
kind='bar', stacked=True, color=colors,
title="Commenting activity and proportion of higher-level comments for {}".format(project).title(),
fontsize=fontsize)
axes.set_ylabel("Number of comments")
axes.legend(bbox_to_anchor=(0.165, 1))
axes2 = axes.twinx()
axes2.set_ylabel("Proportion of Higher-level comments")
axes2.plot(axes.get_xticks(), data['proportion'].values,
linestyle=':', marker='.', markersize=10, linewidth=.7,
color='darkgrey',
fontsize=fontsize)
the_lines = [mlines.Line2D([], [], color='gray', linestyle=':',
marker='.', markersize=10,
label="Proportion")]
axes2.legend(handles=the_lines,
bbox_to_anchor=(1, 1))
ac.show_or_save(show)
def plot_centre_closeness(self, thresh=10, ylim=None, **kwargs):
"""Boxplot of time before return to centre for core authors"""
project, show, _ = ac.handle_kwargs(**kwargs)
timestamps = self.author_frame['timestamps'].apply(np.array)
try:
timestamps.drop("Anonymous", inplace=True)
except ValueError:
pass
delays = timestamps.apply(np.diff)
delays = delays[delays.apply(len) >= thresh]
to_days = np.vectorize(lambda x: x.total_seconds() / (60**2 * 24))
delays = delays.map(to_days)
plt.style.use(SETTINGS['style'])
_, axes = plt.subplots()
bplot = plt.boxplot(delays, sym='.',
showmeans=True, meanline=True)
for key in ['whiskers', 'boxes', 'caps']:
plt.setp(bplot[key], color='steelblue')
plt.setp(bplot['means'], color="firebrick")
axes.set_xticklabels(delays.index, rotation=40, ha='right')
axes.set_xlabel("Participants with at least {} comments".format(
thresh))
axes.set_yticks(np.logspace(-1, 3, num=5, base=2))
axes.set_ylabel("Delay in days")
if ylim:
axes.set_ylim(0, ylim)
axes.set_title("Delays between comments in {}".format(project))
axes.xaxis.set_ticks_position('bottom')
axes.yaxis.set_ticks_position('left')
ac.show_or_save(show)
def scatter_authors(self,
measure="betweenness centrality",
thresh=15, **kwargs):
"""Scatter-plot with position based on interaction and cluster
measure, color based on number of comments, and size on avg comment
length"""
project, show, _ = ac.handle_kwargs(**kwargs)
x_measure, y_measure = [" ".join([netw, measure]) for netw in
["interaction", "cluster"]]
axes = self.author_frame.plot(
kind='scatter',
x=x_measure, y=y_measure,
c='total comments',
s=self.author_frame['word counts'] / self.author_frame[
'total comments'],
cmap="viridis_r",
sharex=False,
title="Author-activity and centrality in {}".format(project))
for name, data in self.author_frame.iterrows():
if data['total comments'] >= thresh:
axes.text(data[x_measure], data[y_measure], name,
fontsize=6)
ac.fake_legend([50, 100, 250], title="Average wordcount of comments")
ac.show_or_save(show)
def plot_centrality_measures(self,
g_type="interaction", measures=None,
delete_on=None, thresh=0, **kwargs):
"""Shows plot of degree_centrality for each author
(only if first measure is non-zero)"""
project, show, fontsize = ac.handle_kwargs(**kwargs)
if not measures:
measures = self.centr_measures
centr_cols, centrality, means = self.__get_centrality_measures(
g_type, measures)
if delete_on is not None:
centrality = centrality[centrality[centr_cols[delete_on]] > thresh]
colors = ac.color_list(len(measures),
SETTINGS['vmin'], SETTINGS['vmax'],
factor=15)
full_measure_names = centrality.columns
centrality.columns = [
col.replace(g_type + " ", "") for col in centrality.columns]
plt.style.use(SETTINGS['style'])
axes = centrality.plot(
kind='bar', color=colors,
title="Centrality-measures for {} ({}-graph)".format(
project, g_type).title())
for measure, color in zip(full_measure_names, colors):
the_mean = means[measure]
axes.lines.append(
mlines.Line2D(
[-.5, len(centrality.index) - .5],
[the_mean, the_mean],
linestyle='-', linewidth=.5,
color=color, zorder=1,
transform=axes.transData))
axes.set_xticklabels(centrality.index, fontsize=fontsize)
ac.show_or_save(show)
def plot_author_activity_bar(self, what='by level', **kwargs):
"""Shows plot of number of comments / wordcount per author.
what can be either 'by level' or 'word counts'"""
project, show, fontsize = ac.handle_kwargs(**kwargs)
plt.style.use(SETTINGS['style'])
if what == "by level":
levels, colors = self.__get_author_activity_bylevel()
total_num_of_comments = int(levels.sum().sum())
axes = levels.plot(
kind='barh', stacked=True, color=colors,
title='Comments per author (total: {})'.format(
total_num_of_comments),
fontsize=fontsize)
axes.set_yticklabels(levels.index, fontsize=fontsize)
elif what == "word counts":
word_counts = self.author_frame[what].sort_values(ascending=False)
total_word_count = int(word_counts.sum())
axes = word_counts.plot(
kind='bar', logy=True,
title='Word-count per author in {} (total: {})'.format(
project, total_word_count),
fontsize=fontsize)
axes.xaxis.set_ticks_position('bottom')
axes.yaxis.set_ticks_position('left')
else:
raise ValueError
ac.show_or_save(show)
def scatter_comments_replies(self, **kwargs):
"""Scatter-plot of comments vs direct replies received"""
project, show, _ = ac.handle_kwargs(**kwargs)
data = self.author_frame[['total comments',
'replies (direct)']]
data.plot(
kind='scatter',
x="total comments", y='replies (direct)',
sharex=False,
title="total comments vs replies in {}".format(project))
ac.show_or_save(show)
def plot_author_activity_hist(self,
what='total comments', bins=10,
**kwargs):
"""Shows plot of histogram of commenting activity.
What can be either 'total comments' (default) or 'word counts'"""
project, show, _ = ac.handle_kwargs(**kwargs)
if what not in set(['total comments', 'word counts']):
raise ValueError
comments = self.author_frame[what]
plt.style.use(SETTINGS['style'])
_, axes = plt.subplots()
comments.hist(bins=bins, grid=False, ax=axes)
axes.set_title("Histogram of {} for {}".format(what, project))
axes.set_xlim(1)
axes.set_yticks(axes.get_yticks()[1:])
ac.show_or_save(show)
def draw_graph(self,
graph_type="interaction",
k=None, reset=False, **kwargs):
"""Draws and shows graph."""
project, show, fontsize = ac.handle_kwargs(**kwargs)
if graph_type == "cluster":
graph = self.c_graph
graph_type = "Co-location Network"
elif graph_type == "interaction":
graph = self.i_graph
graph_type = "Interaction Network"
# attributing widths and colors to edges
edges = graph.edges()
weights = [graph[source][dest]['weight'] * 15 for
source, dest in edges]
edge_colors = [plt.cm.Blues(weight) for weight in weights]
# attributes sizes to nodes
sizes = [(log(self.author_count()[author], 4) + 1) * 300
for author in self.author_frame.index]
# positions with spring
if reset or not self.positions:
self.positions = nx.spring_layout(graph, k=k, scale=1)
# creating title and axes
figure = plt.figure()
figure.suptitle("{} for {}".format(graph_type, project).title(),
fontsize=12)
axes = figure.add_subplot(111)
axes.xaxis.set_ticks([])
axes.yaxis.set_ticks([])
# actual drawing
# consider adding legend
plt.style.use(SETTINGS['style'])
nx.draw_networkx(graph, self.positions,
with_labels=SETTINGS['show_labels_authors'],
font_size=fontsize,
node_size=sizes,
nodelist=self.author_frame.index.tolist(),
node_color=self.author_frame['color'].tolist(),
edges=edges,
width=1,
edge_color=edge_colors,
vmin=SETTINGS['vmin'],
vmax=SETTINGS['vmax'],
cmap=CMAP,
ax=axes)
ac.show_or_save(show)
def plot_activity_degree(self,
g_type='interaction', measures=None,
delete_on=None, thresh=0, **kwargs):
"""Shows plot of number of comments (bar) and network-measures (line)
for all authors with non-null centrality-measure"""
project, show, fontsize = ac.handle_kwargs(**kwargs)
# data for centrality measures
if not measures:
measures = self.centr_measures
if measures == ['hits']:
centr_cols = ['hubs', 'authorities']
centrality = self.__hits()[centr_cols].sort_values(
centr_cols[0], ascending=False)
else:
centr_cols, centrality, _ = self.__get_centrality_measures(
g_type, measures)
if delete_on is not None:
centrality = centrality[centrality[centr_cols[delete_on]] > thresh]
# data for commenting-activity (limited to index of centrality)
comments, colors = self.__get_author_activity_bylevel()
comments = comments.loc[centrality.index]
plt.style.use(SETTINGS['style'])
axes = comments.plot(
kind='bar', stacked=True, color=colors,
title="Commenting activity and {} for {}".format(
", ".join(measures), project).title(),
fontsize=fontsize)
axes.set_ylabel("Number of comments")
axes.xaxis.set_ticks_position('bottom')
axes2 = axes.twinx()
axes2.set_ylabel("Measures")
col_marker = list(zip(centr_cols, "oDsv^"))
for col, marker in col_marker:
axes2.plot(axes.get_xticks(), centrality[col].values,
linestyle=':', marker=marker, markersize=5,
linewidth=.7, color='darkgray')
the_lines = [mlines.Line2D([], [], color='darkgray',
linewidth=.7,
marker=marker,
markersize=5,
label=col.replace(g_type + " ", ""))
for (col, marker) in col_marker]
axes2.legend(handles=the_lines,
bbox_to_anchor=(.83, 1))
ac.show_or_save(show)
def plot_author_activity_pie(self, what='total comments', **kwargs):
"""Shows plot of commenting activity as piechart
what can be either 'total comments' (default) or 'word counts'"""
project, show, fontsize = ac.handle_kwargs(**kwargs)
if what not in set(['total comments', 'word counts']):
raise ValueError
comments = self.author_frame[[what, 'color']].sort_values(
what, ascending=False)
thresh = int(np.ceil(comments[what].sum() / 100))
whatcounted = 'comments' if what == 'total comments' else 'words'
comments.index = [[x if y >= thresh else "fewer than {} {}"
.format(thresh, whatcounted) for
(x, y) in comments[what].items()]]
merged_commenters = comments.index.value_counts()[0]
comments = DataFrame({
'totals': comments[what].groupby(comments.index).sum(),
'maxs': comments[what].groupby(comments.index).max(),
'color': comments['color'].groupby(
comments.index).max()}).sort_values(
'maxs', ascending=False)
for_pie = comments['totals']
for_pie.name = ""
colors = ac.color_list(comments['color'],
SETTINGS['vmin'], SETTINGS['vmax'],
cmap=CMAP)
plt.style.use(SETTINGS['style'])
title = "Activity per author for {}".format(project).title()
if what == "total comments":
title += ' ({} comments, {} with fewer than {} comments)'.format(
int(comments['totals'].sum()),
merged_commenters,
thresh)
else:
title += ' ({} words, {} with fewer than {} words)'.format(
int(comments['totals'].sum()),
merged_commenters,
thresh)
for_pie.plot(
kind='pie', autopct='%.2f %%', figsize=(6, 6),
labels=for_pie.index,
colors=colors,
title=('\n'.join(wrap(title, 60))),
fontsize=fontsize)
ac.show_or_save(show)
def scatter_authors_hits(self, thresh=10, **kwargs):
"""Scatter-plot based on hits-algorithm for hubs and authorities"""
project, show, _ = ac.handle_kwargs(**kwargs)
hits = self.__hits()
axes = hits.plot(
kind='scatter',
x='hubs', y='authorities',
c='total comments',
s=hits['word counts'] / hits['total comments'],
cmap="viridis_r",
sharex=False,
title="Hubs and Authorities in {}".format(project))
for name, data in hits.iterrows():
if data['total comments'] >= thresh:
axes.text(data['hubs'], data['authorities'], name,
fontsize=6)
ac.fake_legend([50, 100, 250], title="Average wordcount of comments")
ac.show_or_save(show)
def plot_centre_crowd(self, thresh=2, show_threads=False, **kwargs):
"""Plotting evolution of number of participants close to centre"""
project, show, _ = ac.handle_kwargs(**kwargs)
data = self.__get_centre_distances(thresh, split=False)
data_close = DataFrame({
'6 hours': data[data <= .25].count(axis=1),
'12 hours': data[(data <= .5) & (data > .25)].count(axis=1),
'24 hours': data[(data <= 1) & (data > .5)].count(axis=1)},
columns=['6 hours', '12 hours', '24 hours'])
plt.style.use(SETTINGS['style'])
y_max = data_close.sum(axis=1).max()
_, axes = plt.subplots()
data_close.plot(kind="area", ax=axes, stacked=True,
color=['darkslategray', 'steelblue', 'lightgray'])
axes.set_yticks(range(1, y_max + 1))
axes.set_ylabel("Number of participants")
axes.set_title("Crowd close to the centre of discussion in {}".format(
project))
axes.xaxis.set_ticks_position('bottom')
axes.yaxis.set_ticks_position('left')
if show_threads:
self.__show_threads(axes)
ac.show_or_save(show)
def plot_i_trajectories(self,
thresh=None, select=None, l_thresh=5,
**kwargs):
"""Plots interaction-trajectories for each pair of contributors."""
project, show, _ = ac.handle_kwargs(**kwargs)
trajectories = {}
for (source, dest, data) in self.i_graph.edges_iter(data=True):
name = " / ".join([source, dest])
trajectories[name] = Series(Counter(data['timestamps']),
name=name)
try:
tr_data = DataFrame(trajectories)
except ValueError as err:
print("Could not create DataFrame: ", err)
else:
tr_data = tr_data.fillna(0).cumsum().sort_index()
col_order = tr_data.iloc[-1].sort_values(ascending=False).index
tr_data = tr_data[col_order]
title = "Interaction trajectories for {}".format(project)
if select:
tr_data = tr_data.iloc[:, :select]
title += " ({} largest)".format(select)
elif thresh:
tr_data = tr_data.loc[:, ~(tr_data < thresh).all(axis=0)]
title += " (minimally {} interactions)".format(thresh)
plt.style.use(SETTINGS['style'])
_, axes = plt.subplots()
for col in col_order[:l_thresh]:
tr_data[col].plot(ax=axes, label=col)
for col in col_order[l_thresh:]:
tr_data[col].plot(ax=axes, label=None)
axes.legend(labels=col_order[:l_thresh], loc='best')
axes.set_title("Interaction trajectories for {}".format(project))
axes.xaxis.set_ticks_position('bottom')
axes.yaxis.set_ticks_position('left')
ac.show_or_save(show)
def draw_centre_discussion(self,
regular_intervals=False,
skips=2, zoom=12, **kwargs):
"""Draws part of nx.DiGraph to picture who's
at the centre of activity"""
_, show, _ = ac.handle_kwargs(**kwargs)
activity_df = self.author_frame[
['color', 'angle', 'timestamps']].copy()
if not regular_intervals:
intervals = np.concatenate(activity_df['timestamps'].values)
intervals.sort(kind='mergesort')
intervals = intervals[::skips]
else:
start = np.min(activity_df['timestamps'].apply(np.min))
stop = np.max(activity_df['timestamps'].apply(np.max))
intervals = date_range(start, stop)[::skips]
x_max, y_max = 0, 0
for interval in intervals:
interval_data = activity_df['timestamps'].apply(
# TODO: fix - does not work
lambda x, intv=interval: x[x <= intv])
try:
interval_data = interval_data.apply(
lambda x, intv=interval: (intv - x[-1]).total_seconds()
if x.size else np.nan)
except AttributeError:
interval_data = interval_data.apply(
lambda x, intv=interval:
(intv - x[-1]) / np.timedelta64(1, 's')
if x.size else np.nan)
x_coord = interval_data * np.cos(activity_df['angle'])
the_min, the_max = np.min(x_coord), np.max(x_coord)
x_max = max(abs(the_max), abs(the_min), x_max)
y_coord = interval_data * np.sin(activity_df['angle'])
the_min, the_max = np.min(y_coord), np.max(y_coord)
y_max = max(abs(the_max), abs(the_min), y_max)
coords = DataFrame({"x": x_coord, "y": y_coord})
assert interval not in activity_df.columns
activity_df[interval] = [list(x) for x in coords.values]
in_secs = {'day': 86400, '2 days': 172800, 'week': 604800,
'1 week': 604800, '2 weeks': 1209600, '3 weeks': 1814400,
'month': 2635200}
try:
xy_max = max(x_max, y_max) / zoom
except TypeError:
xy_max = in_secs[zoom]
except KeyError:
xy_max = max(x_max, y_max)
def get_fig(activity_df, col_name):
"""Helper-function returning a fig based on DataFrame and col."""
plt.style.use(SETTINGS['style'])
coord = activity_df[col_name].to_dict()
dists = pd.DataFrame(activity_df[col_name].tolist(),
columns=['x', 'y'],
index=activity_df.index)
dists = np.sqrt(dists['x']**2 + dists['y']**2)
in_day = dists[dists < in_secs['day']].count()
in_week = dists[dists < in_secs['week']].count()
in_month = dists[dists < in_secs['month']].count()
fig = plt.figure()
# left plot (full)
axes1 = fig.add_subplot(121, aspect='equal')
axes1.set_xlim([-xy_max, xy_max])
axes1.set_ylim([-xy_max, xy_max])
axes1.xaxis.set_ticks([])
axes1.yaxis.set_ticks([])
day = plt.Circle((0, 0), in_secs['day'], color='darkslategray')
week = plt.Circle((0, 0), in_secs['week'], color='slategray')
month = plt.Circle((0, 0), in_secs['month'], color="lightblue")
axes1.add_artist(month)
axes1.add_artist(week)
axes1.add_artist(day)
the_date = pd.to_datetime(str(col_name)).strftime(
'%Y.%m.%d\n%H:%M')
axes1.text(-xy_max / 1.07, xy_max / 1.30, the_date,
bbox=dict(facecolor='slategray', alpha=0.5))
# axes.text(in_secs['day'], -100, '1 day', fontsize=10)
# axes.text(in_secs['week'], -100, '1 week', fontsize=10)
# axes.text(in_secs['month'], -100, '1 month', fontsize=10)
nx.draw_networkx_nodes(self.graph, coord,
nodelist=activity_df.index.tolist(),
node_color=activity_df['color'],
node_size=20,
cmap=CMAP,
ax=axes1)
# right plot: zoomed
axes2 = fig.add_subplot(122, aspect='equal')
axes2.set_xlim([-xy_max / 10, xy_max / 10])
axes2.set_ylim([-xy_max / 10, xy_max / 10])
axes2.xaxis.set_ticks([])
axes2.yaxis.set_ticks([])
day = plt.Circle((0, 0), in_secs['day'], color='darkslategray')
week = plt.Circle((0, 0), in_secs['week'], color='slategray')
month = plt.Circle((0, 0), in_secs['month'], color="lightblue")
axes2.add_artist(month)
axes2.add_artist(week)
axes2.add_artist(day)
day_patch = mpatches.Patch(
color='darkslategray',
label="{: <3} active in last day".format(in_day).ljust(25))
week_patch = mpatches.Patch(
color='slategray',
label="{: <3} active in last week".format(in_week).ljust(25))
month_patch = mpatches.Patch(
color='lightblue',
label="{: <3} active in last month".format(in_month).ljust(25))
plt.legend(handles=[day_patch, week_patch, month_patch])
nx.draw_networkx_nodes(self.graph, coord,
nodelist=activity_df.index.tolist(),
node_color=activity_df['color'],
node_size=20,
cmap=CMAP,
ax=axes2)
return fig
ion()
for (num, interval) in enumerate(intervals):
fig = get_fig(activity_df, interval)
if show:
fig.canvas.draw()
plt.draw()
plt.pause(1)
plt.close(fig)
else:
plt.savefig("FIGS/img{0:0>5}.png".format(num))
plt.close(fig)