Time Machine Status Update 5

I give another status update about the time-traveling debugger with Huiqi. We slipped into a little bit of dynamic programming talk in the middle of the conversation. Enjoy!

Transcript

The following transcript was automatically generated by an algorithm.

00:00 : so
00:01 : remember i had this code panel and also
00:04 : the timeline panel and
00:06 : now now you can also like skip around in
00:09 : the timeline you can click on
00:11 : the
00:12 : place in the timeline it'll jump to
00:15 : there
00:16 : and also use the arrows
00:20 : and i added this view menu
00:23 : that can let you
00:25 : that can let you like like if if your
00:27 : panel got hidden by another panel
00:30 : then you can say bring it to the top
00:34 : and
00:35 : and also if you
00:36 : close the panel
00:38 : then you can always bring it back
00:42 : okay yeah will there be like uh auto
00:46 : arrangement
00:47 : it's like putting it in a place so that
00:50 : it won't like overlap with each other
00:53 : yes
00:54 : i plan to do that at some point yes okay
00:59 : yep yeah yes
01:01 : that should definitely be there but
01:04 : uh
01:05 : but i uh but there's other fish to fry
01:08 : for now um
01:10 : so
01:11 : and then other than that the one thing i
01:13 : actually spent a lot of time on is
01:16 : actually the c code
01:18 : program
01:19 : oh not the c code program oh you know
01:22 : the c code program which is for um
01:26 : like displaying the timeline on the
01:29 : console
01:31 : like that
01:32 : okay um
01:35 : let's do this fibonacci program actually
01:44 : okay
01:46 : so okay that's a fibonacci program that
01:49 : had to calculate a big
01:51 : number probably
01:53 : i don't know how big oh i can
01:57 : there it is
01:59 : of 18 so it's pretty big
02:02 : um
02:05 : i made another one called c calls
02:09 : which gives you
02:13 : which gives you the
02:14 : summary
02:16 : uh in terms of function calls so this is
02:19 : more summarized version of that okay
02:22 : where
02:24 : well that's taking a while where did uh
02:27 : well maybe i should just do a smaller
02:30 : one
02:39 : and i also had to spend a bit of time
02:44 : uh optimizing it because it was really
02:47 : slow initially and and now it's only
02:50 : somewhat slow
02:52 : um
02:53 : okay so so you you get this call graph
02:57 : which which basically i stole from uh
03:01 : bill lewis's
03:03 : odb
03:04 : omniscient debugger he
03:06 : he also has a panel in his java based ui
03:11 : that is a tree like this
03:13 : yeah and i basically i call this the
03:15 : outline
03:17 : um of all the calls
03:19 : and it'll show you the
03:22 : input parameters and the output of every
03:25 : call so it's it's just a very nice
03:27 : summary of what happened in your program
03:30 : in terms of function calls
03:34 : and
03:36 : yeah so
03:37 : i think
03:38 : for for like more like a functional
03:41 : programming approach like if all your
03:43 : program is
03:45 : stuff didn't this one this won't help
03:48 : yep that's true yeah that's kind of true
03:51 : um
03:52 : yeah i agree like if you if you if you
03:55 : only have one function call
03:57 : in the whole program
03:59 : and then the program is just doing a lot
04:01 : of loops and
04:03 : yeah
04:04 : stuff then this won't show you anything
04:07 : at all
04:08 : um but yeah uh
04:10 : but i think this
04:12 : can this i think would be really nice
04:16 : for
04:18 : teaching algorithms actually
04:21 : um
04:24 : like like i've actually been planning
04:27 : for a talk
04:29 : with insight and on dynamic programming
04:33 : and
04:34 : and one of the
04:36 : there's two ways to do dynamic
04:38 : programming one of them is with
04:40 : recursion but which you can use this and
04:44 : then you can use this graph to show
04:48 : uh why
04:49 : this algorithm is very inefficient
04:52 : there's a there's a and
04:54 : there's an mit talk
04:56 : that uh
04:58 : that explains dynamic programming in
05:00 : terms of the fibonacci sequence at this
05:03 : also because i think it's a good example
05:06 : to show
05:07 : because
05:08 : fibonacci sequence gets taught so many
05:11 : times that people probably already know
05:13 : it
05:14 : and
05:16 : and so you can kind of go faster because
05:20 : you can probably in that mit course that
05:23 : guy said
05:25 : hey
05:26 : and this algorithm is very inefficient
05:30 : because
05:31 : and in order to explain why it's
05:33 : inefficient he he had to use these math
05:37 : proofs and these these theorems that
05:40 : they established earlier in the class
05:44 : and
05:44 : use this theorem to show that
05:48 : because there's recursion the recursion
05:50 : is done this way therefore
05:54 : the
05:55 : the program has to take
05:57 : this long
05:58 : you know related to the size of the
06:00 : input and therefore it's an inefficient
06:02 : algorithm
06:03 : but i think
06:05 : if i have this tool
06:07 : all i have to do is first of all show
06:09 : them this guy is identical to this guy
06:14 : why are we redoing that calculation and
06:17 : and not only that that guy
06:22 : that guy
06:25 : and that guy
06:27 : those are identical why are we oh that
06:30 : happened yet again
06:33 : like that's ridiculous
06:35 : and not so like
06:37 : this
06:39 : that we're we're just doing a lot of
06:41 : repeated work so we don't need to do
06:43 : that and you can very easily show that
06:47 : here
06:49 : okay
06:50 : so why it's called dynamic programming
06:53 : it while what you are doing is just you
06:55 : are basically recording what it has been
06:59 : done before and just use it don't worry
07:01 : about the name so the mit guy actually
07:04 : explained
07:05 : why it's called dynamic programming
07:07 : basically
07:08 : they wanted a name that sounds
07:11 : like it's not math
07:14 : because that research center they were
07:18 : they didn't like people working on math
07:20 : or something
07:22 : i don't know they just it was not a good
07:24 : name and they purposefully made it not a
07:27 : good name so the name doesn't really
07:29 : matter
07:30 : okay but basically like once you use the
07:33 : caching
07:34 : then this happens right and that's
07:37 : that's an extremely good illustration of
07:41 : that the point of
07:43 : what it is and and
07:45 : essentially is
07:46 : recursion with caching or recursion with
07:50 : memoization is what they call it
07:52 : uh there's another way there's a
07:54 : bottom-up way to do it but you know
07:57 : whatever i i'm not that's not the point
08:00 : we don't have to get into it my point is
08:02 : this
08:03 : i think this is a really good tool for
08:05 : teaching algorithms
08:07 : um
08:08 : so
08:09 : i plan to use it for that and like
08:12 : that's one of the goals of this project
08:15 : i'm happy to say the debugger has been
08:18 : holding up well even though not not
08:21 : every bug is easy to solve yet
08:25 : it still takes
08:26 : concentration to
08:29 : navigate through the code to the place
08:32 : where
08:34 : the problem is
08:36 : but at least it's straightforward like
08:38 : it's somewhat it it requires you to
08:41 : concentrate and it's a little bit
08:42 : tedious to drill into the code navigate
08:46 : go either like step over or step inside
08:49 : or whatever
08:52 : to find the code but i think
08:55 : uh i think with
08:56 : the features like searching through the
09:00 : database and
09:02 : as easy feature would just be switching
09:04 : to another source file
09:07 : uh there's isn't that feature doesn't
09:09 : exist right now like if i can just just
09:11 : pick which source file i want to go to
09:14 : and then click on one of the lines
09:17 : to jump to that line
09:20 : that would already help a lot with the
09:22 : navigation
09:23 : but also
09:25 : also the object timeline thing would
09:27 : really really help as in if you see that
09:30 : object having a weird value
09:33 : it just go back in time to last time it
09:36 : was modified or or maybe go back to
09:40 : before it that had had that value
09:44 : that
09:45 : yeah i know that's very doable
09:48 : i just need to do it
09:51 : okay it would be like a separate panel
09:54 : for
09:55 : probably separate panel um
09:58 : yeah
09:59 : so so we'll see we'll see
10:02 : okay
10:03 : um i'm basically searching for
10:07 : searching for the ideal debugging
10:09 : experience
10:11 : by about trial and error right now
10:14 : probably i was thinking at one point
10:17 : when i think this is ready i might go to
10:20 : university
10:22 : and ask people to learn it and then i'll
10:25 : test them like they'll do an experiment
10:28 : on the people like have a have a control
10:31 : group and
10:32 : and an experimental group
10:34 : and people who debug
10:37 : nor using their usual way
10:40 : versus
10:41 : versus people who have learned to use
10:43 : this debugger yeah i think it would be
10:46 : cool if the university can find you that
10:48 : would be nice and then like you can
10:50 : publish a paper
10:52 : about that
10:54 : yeah um
10:55 : the problem is
10:57 : i won't get paid that much at all it'll
10:59 : be a very very severe pay cut
11:03 : for
11:04 : what i'm used to
11:05 : if you're
11:06 : established as a researcher
11:09 : or you're really good at getting funding
11:12 : you can get
11:14 : paid pretty well if you're good at
11:16 : writing grant proposals
11:19 : and you can get the student to work for
11:21 : you
11:22 : like
11:23 : like uh like having a research assistant
11:26 : that kind of stuff sure yeah that's
11:30 : that's possible oh yeah
11:32 : i have talked to my my former professor
11:35 : he asked me to get a phd