def process_one_item(table):
monitor.incr("process."+table)
# grab the most recent issue to process
[issue_id, d, t1] = claim_next(table)
print(table, issue_id, d)
# get the text
[text, t2] = get_text(table, issue_id)
# score the text
[score_report, t3] = toxicity_report.compute_prediction_report(text)
result = {"toxicity."+VERSION: score_report}
# write results to db
t4=update_db(table,issue_id,result)
monitor.timing("db.next."+table,int(t1*1000))
monitor.timing("db.gettext."+table,int(t2*1000))
monitor.timing("db.writeresult."+table,int(t4*1000))
monitor.timing("scoring."+table,int(t3*1000))
monitor.gauge("lastprocessed."+table,parse(d).timestamp())
if score_report["score"]==1:
monitor.incr("foundtoxic."+table)
def test_toxic():
[report,
time] = toxicity_report.compute_prediction_report("Kill this process!")
assert report["score"] == 1
assert report["alt_tried"] == 2
def test_code():
# should behave like empty
[report, time] = toxicity_report.compute_prediction_report("`hate`")
assert report["score"] == 0
assert report["reason"] == "empty"
def test_nontoxic():
[report, time] = toxicity_report.compute_prediction_report("test")
assert report["score"] == 0
assert report["en"] == 1
assert report["orig"]["persp_raw"]["detectedLanguages"] == ["en"]
def test_empty():
[report, time] = toxicity_report.compute_prediction_report("")
assert report["score"] == 0
assert report["reason"] == "empty"