[tor-bugs] #19259 [Metrics/Onionoo]: uncaught NFE

[Tor Bug Tracker & Wiki]

#19259: uncaught NFE
-----------------------------+-----------------------------------
 Reporter:  iwakeh           |          Owner:  iwakeh
     Type:  defect           |         Status:  needs_information
 Priority:  High             |      Milestone:
Component:  Metrics/Onionoo  |        Version:
 Severity:  Major            |     Resolution:
 Keywords:                   |  Actual Points:
Parent ID:                   |         Points:
 Reviewer:                   |        Sponsor:
-----------------------------+-----------------------------------

Comment (by karsten):

 Oh, sorry for misunderstanding.  Let me reply to this comment instead:

 Replying to [comment:10 iwakeh]:
 > Test class commentary
 >
 > `testSetup` for warm-up, passes

 Sounds good. :)

 > `testWeightHistory` fails, document contains `NaN` after compression

 It seems there are two problems with this test case:

  1. Intervals should be at least one hour long for the internal math to
 work.  This is unfortunate, though the assumption is realistic for real
 data.  But let's relax this requirement and accept intervals shorter than
 1 hour.  A plausible minimum would be 1 second, because we cannot
 represent anything shorter than that.
  1. Entries are only compressed if they are a given number of days in the
 past, calculated from the current system time.  A better approach would be
 to compress based on the end time of the most recent history entry to
 remove yet another dependency on system time.  For the test to work you'll
 have to date entries back by 1 year or maybe even more.

 > `testSetFromDocumentString` passes

 Okay.

 > `testMissingValues` fails, b/c the small negative values are not
 recognized as missing

 How about we represent missing values on disk as empty string (`""`),
 which also saves disk space, and in memory as `-1.0` to keep the current
 code working?

 > `testSetFromDocumentStringWrongInput` fails, error b/c of nfe

 Right, let's fix that nfe.

 > `testSetFromDocumentStringWrongDateTime` failes, b/c line with later
 beginning than end is not rejected.

 Right, let's reject those lines.  That means additional tests, but those
 are cheap and might make the code more robust.

 > `testCompare` fails, b/c there is an entry with the key from the first
 and values from the second element.

 Let's reject those lines, too.

 Is that more what you had in mind?

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/19259#comment:13>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online