def main(tabfile, dgraph):
""" (-> Path-String GraphDict Result)
Input: a .tab file, an overall summary of running all configurations
of a project.
Output: a Result object.
"""
print("Collecting results from ground truth data '%s'" % tabfile)
fname = util.strip_suffix(tabfile).rsplit("/", 1)[-1]
num_modules = count_modules(tabfile)
num_configs = 2 ** num_modules
print("Project contains %s modules (%s configurations)" % (num_modules, num_configs))
u_raw = all_cells_matching(tabfile, config.is_untyped)
g_raw = all_cells_matching(tabfile, lambda x: not (config.is_typed(x) or config.is_untyped(x)))
t_raw = all_cells_matching(tabfile, config.is_typed)
ugt_violin = plot.violin([u_raw, g_raw, t_raw]
,"%s_untyped-vs-gradual-vs-typed" % fname
,"Configuration"
,"Runtime (ms)"
,xlabels=["untyped","gradual\n(all configs)","typed"]) if u_raw and g_raw and t_raw else None
# Collect absolute BEST and WORST times+configs
ten_percent = max(3, min(10, int(0.10 * num_configs)))
best_cfg_and_times = best_rows(tabfile
,ten_percent
,lambda acc,tmp: tmp[1] < acc[1]
,lambda row:(row[0], int(statistics.mean([int(x) for x in row[1::]]))))
worst_cfg_and_times = best_rows(tabfile
,ten_percent
,lambda acc,tmp: acc[1] < tmp[1]
,lambda row:(row[0], int(statistics.mean([int(x) for x in row[1::]]))))
stats = {
"title" : fname
,"runs" : count_runs(tabfile)
,"graph" : dgraph
,"ugt" : {"img" : ugt_violin
,"summary" : {"untyped" : basic_row_stats(u_raw)
,"gradual" : basic_row_stats(g_raw)
,"typed" : basic_row_stats(t_raw)}}
,"best" : [config.basic_stats(v[0], v[1], dgraph)
for v in best_cfg_and_times if v is not None]
,"worst" : [config.basic_stats(v[0], v[1], dgraph)
for v in worst_cfg_and_times if v is not None]
,"bucketed": plot.violin(bucketize(tabfile, config.num_typed_modules, num_modules)
,"%s_by-typed-modules" % fname
,"Number of Typed Modules"
,"Runtime (ms)"
,positions=range(0, 1+num_modules)
,xlabels=range(0, 1+num_modules))
,"fixed" : [plot.double_violin([bucketize(tabfile, config.num_typed_modules, num_modules, pred=lambda cfg: config.untyped_at(cfg, v[0]))
,bucketize(tabfile, config.num_typed_modules, num_modules, pred=lambda cfg: config.typed_at(cfg, v[0]))]
,"%s_fixing-%s" % (fname, k)
,"Number of Typed Modules"
,"Runtime (ms)"
,legend=["%s is untyped" % k
,"%s is typed" % k])
for (k,v) in sorted(dgraph.items(), key=lambda item:item[1][0])]
}
return stats
def graph_absolute_runtimes(self, preds, xtitle, xlabels, output, title=None):
"""
Plot the absolute running times of the configurations picked by
each predicate in `preds`.
(The number of columns in the result is equal to the size of `preds`)
"""
columns = [self.stats_of_predicate(pred)["raw"]
for pred in preds]
# TODO filter empty columns ?
return plot.violin(columns
,title or "%s absolute runtimes" % self.get_project_name()
,xtitle
,"Runtime (ms)"
,xlabels=xlabels
,output="%s/%s" % (self.output_dir, output))
