[tor-bugs] #14011 [Stem]: Implement lazy parsing for stem

[Tor Bug Tracker & Wiki]

#14011: Implement lazy parsing for stem
-----------------------------+--------------------------------
     Reporter:  phw          |      Owner:  phw
         Type:  enhancement  |     Status:  new
     Priority:  major        |  Milestone:
    Component:  Stem         |    Version:
   Resolution:               |   Keywords:  descriptor, zoossh
Actual Points:               |  Parent ID:
       Points:               |
-----------------------------+--------------------------------

Comment (by atagar):

 What does "ns/op" stand for? For benchmarking I plan to point Stem at some
 CollecTor archives (https://collector.torproject.org/) to get
 descriptors/s.

 I've been working on this for Stem over the last couple weeks and very,
 very close to being done (finished with everything except the router
 status entries). I'll be sending an email this weekend with the results.

 >  As expected, lazy parsing is much faster when the content of router
 statuses or descriptors is not accessed. However, when the content of all
 statuses/descriptors is accessed, lazy parsing is slower as it involves
 more overhead.

 Gotcha. For Stem lazy loading means a little more memory usage (since
 we're storing an extra 'keyword => values' dict) but I wouldn't expect
 accessing all attributes to be appreciably slower since it's pretty much
 an identical codepath as eager loading.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/14011#comment:4>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online