During my batch I would post regular updates on Zulip about what I had been working on. Over the course of my batch this became a nice timeline of sorts. I am publishing it here in case anyone is curious.
First checkin here!
Progress so far
Researched a bunch of resources as a starting point for my load
balancer project. Worked through
Load balancing and the power of hashing,
a nice blog post. Turned out to have some serious math, deeper than I
@Katrina Evtimova (S1'17) helped out a lot throughout
yesterday and today. Thank you so much!
- Read through other resources on load balancing, which includes a few papers.
- Start working on getting more familiar with Erlang, which is what I want to use for this project
- Maintain a healthy balance between 1 & 2
P.S.: Feels nice to have been able to checkin finally with something concrete.
P.P.S: The blog has an amazing collection of posts related to Math & CS combined together. Anyone interested should totally check it out.
Yesterday and day before
- Spoke to
@Nicole Orchard (SP2'17)about bikes in depth. And picked up quite a few tips on how to buy one. Thank you so much!
- Discussed my project ideas with
@Saul Pwanson (SP2'17). The load balancer seems a bit abstract at the moment. It is a better idea to start off with a UNIX shell instead, considering that I want to be doing systems programming here.
- Started reading up about shell, and wrote about 4 lines of code in C.
@Joe Ardent (SP1'17)helped over Zulip for the linker errors I was facing on OSX.
- Make some real progress on the UNIX shell
- Pair with others on C, Python or anything else.
Friday and weekend
- Paired with
@Jaseem Abid (S1'17)on LLVM code generation for a block of C. Turned out to be a fun and educating exercise. Will do this more here.
- Read and understood about fork better than I thought I did. Thanks
@Saul Pwanson (SP2'17)for helping me understand the concepts. Wrote a blog post on that, which seems to have gone viral on Hackernews.
- Discussed about the slab allocator with
@Tim Babb (S1'17)and
@Jaseem Abid (S1'17). I was the lurker for most parts of the conversation.
- Wrote another blog post on week 1 of RC.
Today and the week ahead
- Implement command execution for the shell that I am building
- Pair with
@Tim Babb (S1'17)&
@Jaseem Abid (S1'17)on the slab allocator sometime this week
- Join the Haskell group. Work through the first five chapters of the Haskell book before the meetup
- Work through the first chapter of Programming Pearls by Jon Bentley. This will be a weekly exercise. If anyone wants to join, you are most welcome.
- Realized that a language like Python makes you forget the nitty gritty details that goes on under the hood.
- Implemented the command line parser for the shell, which is
string.strip().split(" ")in Python.
- Learned about dynamic memory allocation in C (more like recalled the forgotten)
- Ran into seg faults. Fixed them
- Get command execution to work. Currently facing another seg fault (no surprises there!)
- Read up on slab allocator so that I can contribute on that more effectively while pairing
- Attend “Best Programming Talks at 7pm”
- Fixed bugs in the parser.
- Paired with
@Dominic Michael Spadacene (SP2'17)on the shell. Fixed bugs and cleaned up the code. We now have an idea of how much memory this will use. Brainstormed on why
cdwouldn’t work out of the box like
pwd. Turns out it is a shell builtin, and we’ll have to implement this on our own. To be done later.
- Pushed to github: https://github.com/indradhanush/rc-shell
- Paired with
@Jaseem Abid (S1'17)on improving the C functions to accept pointers instead. Code started seg faulting again. Used
valgrindto see if program is leaking memory. Turns out it is leaking quite a bit. Will get back to this later in the week.
- Attended the Best Programmin Talks event. Watched
States and Nomads.
- Take a break from the shell.
- Pair with
@Saul Pwanson (SP2'17)on visidata.
- Start going through
the Haskell Bookand
Learn you a Haskell for Great Goodto be on the same page as everyone for the Haskell event later this week.
- Do whatever it takes to arrive at RC earlier tomorrow onwards. Have been reaching late for the last 2 days.
- Paired with
@Saul Pwanson (SP2'17)to implement Github API as a browsable sheet. We also ended up with a generic client design that can be reused for other APIs.
- Read through the first chapter of
Learn You a Haskell for Great Goodand mid way through the first chapter on lambda calculus of the
- Reached RC earlier than before. Yay! Would like to keep pushing this towards 9 AM.
- Read more of Haskell Book
- Prepare for a presentation
- Maybe read up on some C if time permits
- Read more of “Learn You a Haskell for Great Good”. Specifically, I am familiar with Haskell’s syntax, declaring variables, functions with type declarations (is that the correct term?), lists, strings, list comprehensions, tuples, type classes and type annotations.
- Finally felt like I understood Lambda Calculus, thanks to
@Tim Babb (S1'17),
@Stefan Lesser (S1'17)and
@Jaseem Abid (S1'17)multiple times.
- Prepared and gave a talk at RC presentations on “Getting noticed on the Internet and what to do about it”.
@Vaibhav Sagar (F1'16)’s advice on leaving the laptop at RC. This sounds like a great thing to do, plus my backpack is much lighter.
- Managed to bring down my commute time on the bike from ~50 minutes to ~40 minutes, mostly because I’m now more used to the ride and the route. Bringing this further down to ~35 minutes is a challenge.
- Publish the transcript of my talk from yesterday as a blog post
- Read about pointers in C in depth and refactor the existing code of the shell I was working on.
- Read about
PTS/TTYand signal handling and start implementing it in the shell
- Published Getting noticed on the Internet and what to do about it
- Read up about pointers. Refactored my shell to use pointers.
- The refactoring added some bugs; Fixing them required me to learn
gdb. Thanks to both
@Trevor Saunders (S1'17)and
@Sabelo Mhlambi (S1'17)for helping me get familiar with gdb. This turned out to be much faster, than digging through the manual myself.
- More bugs. Got sucked into a rabbit hole. Reminds me of this poem:
- 99 little bugs to fix, 99 little bugs to fix, take one out, and patch it around, 117 little bugs to fix!
@Saul Pwanson (SP2'17)suggested to use
valgrindfor one of the weird segmentation fault related bugs. Thank you! Using
valgrindis also faster than using
printffor debugging. Will use more of this.
- Attended the Haskell study group meet up. Suggested ways to proceed further by maybe forming groups where people have similar projects in mind to implement in Haskell. This could be useful for people for whom Haskell is the side project and not the priority. That would help save time for everyone in this group while making some pet project out of Haskell.
- Slogged a bit (It was a long week, but it came and went, and I didn’t realise)
- Started reading
Linux Systems Programmingby Robert Love. Read through the introduction chapter which talks about linux fundamentals. I knew a few things already, while learned a few new things along the way.
- Read about signal handling as well, and started playing around. More reading on signal handling is required before I can incorporate it into my shell.
- Had my mind blown over at the American Museum of Natural History
- Wrote a blog post - Exploration and Prioritisation - Week two of RC
- Read up on signal handling from Linux System Programming and implement it in the shell
- Paired with
@Dominic Michael Spadacene (SP2'17)to implement
Control-Chandling in the shell. Came up with a hacky way to do this such that
Control-Cdoes not kill the shell, but only the command running in the shell instead.
- Wasn’t happy with hack. Researched for other ways and spoke to
@Saul Pwanson (SP2'17)about this. That’s when we realised that the hack was probably the way to implement it, unless I want to dig into
pts/ttyhandling. Opening a pts should make handling
- Started reading up on pts.
- More reading up on
ptsand implement it in the shell. This sounds like fun and painful at the same time.
- Dug deep to read through the chapters on Process control and Signals
Advanced Programming in the UNIX Environment
- Figured out a clean solution of handling keyboard interrupt. We can
sigactionto get the parent to ignore the interrupt and set it as the default action on the child process. This makes me happy because it is not a hack anymore.
- Explained this to
@Saul Pwanson (SP2'17)and
@Dominic Michael Spadacene(SP2’17) one by one. This helped me to find the blind spots in my own understanding. I digged into them and cleared those as well. Writing a blog post will reveal if there are more of those.
- Realised that generally a typical day at RC rarely goes as planned. Which seems to be a nice thing.
- Overall feels like a productive day even though I wrote < 20 lines of code.
- Stop using printf in the signal handler since it is not async-safe. Read up on alternatives and implement that.
- Start implementing job control. I will have to implement the
fgas builtin commands as a part of this exercise
- Go to the concert at Prospect park in the evening.
- Completed the clean handling of Ctrl-C in the shell
- Started working on supporting atleast one builtin command
- Attended Beginner’s Club: Design Patterns by
@Andrea Law (SP2'17); An interesting discussion followed. I digressed a bit from the topic and spoke about McCabe’s complexity and why everyone should use it right from the start of the project. Connected it back to the topic by saying that by keeping your code complexity low from the start, it can help to pick simpler design patterns instead of using something that sounds cool but is probably an overkill for a problem
- Attended session on taking care of our hands by
@Naomi Saphra (SP2'17); Really helpful that in case I ever develop symptoms I’ll be able to notice them before it gets bad.
- Had a fun time at the prospect park concert
- Finish builtins
- Start working on job control
- Implemented the builtin command
cd; The code for this is very adhoc and could definitely do with some improvements. I will get back to this later.
- Was struggling with breaking the single
main.cinto a modular structure. Paired with
@Trent Gill (SP2'17), who explained the ideas behind
.cfiles. Blog posts on the internet and answers on stack overflow on this were conflicting. Also improved the
Makefilealong the way.
- The shell can now run background commands if you append an
&at the end of a command. There is no job control yet.
- Fixed memory leaks and bugs that kept cropping up.
- More code cleanup
@Tim Babb (S1'17)’s session on Automatic Differentiation. Key takeaway: Just because something has a fancy name doesn’t mean it is rocket science. It turned out to be very simple instead.
- It was the first day where I wrote code for most part of the day instead of reading. I feel happy again.
- Take a break from the shell for now
- Mock algorithm interview
- Work through first chapter of Programming Pearls.
- Worked through the first two chapters of Programming Pearls. Also thought about few of the problems given at the end of each chapter.
- Gave a mock interview on algorithms. Realised how rusty I’ve become on that.
- Worked through the first chapter and its exercises of Algorithms by S.Dasgupta
- Got myself an NYPL card
- More algorithms
Saturday & Sunday
- Read through a little more of Algorithms by S.Dasgupta
- Finished reading
Distributed Systems for Fun and Profit. This was lying half read for quite some time. Key takeaway: It’s a good book if you want to get started in distributed systems but don’t know how. The book touches on all the topics in a light manner without going into too much detail. Now I know what things I need to pick up and start digging deeper into, and most importantly in what order. Recommended reading for beginners in this field.
- Wrote and published Writing a UNIX shell - Part II ; Polishing a blog post is a lot of more work. Phew.
- Implement job control ( the shell can already run background tasks
commands if you suffix them with
- Got sucked into the world of pts in my quest to understand job control. I probably don’t need this yet. On the positive side, I think I probably understand what I need now.
- Refactored the code for builtins with a cleaner interface with
@Saul Pwanson (SP2'17)
- Conquer job control (hopefully)
@Seth Schori (SP2'17)with a circular import issue he was facing on his flask app.
- Spoke to
@Andrea Law (SP2'17)about docker. She was stuck on the exercises from a docker tutorial because they were for linux, and she’s on OSX instead. Showed her how she could spin up a docker container with linux to work on the tutorial instead.
- Some reading and tinkering with job control. Overall not a very productive day.
- Fat Cat game night. That was fun!
- Late start to the day. Skipping RC in favour of working out of NYPL instead.
- Was able to make some headway in terms of code for job control
- Finished the implementation of job control on the shell. Processes
clean up after themselves if they run in the background or are
interrupted in between. Was stuck on this for quite some time and
@Saul Pwanson (SP2'17)to help me on this. Thank you so much!
- The shell now runs directly under a terminal emulator, instead of having to first open bash and then run the executable. This is the biggest win after struggling with job control for all of this week.
- Haven’t felt this happy about making something in a long time.
- Discovered an interesting bug regarding the handling of SIGTTOU. Need to explore this. Also will implement the job completion message tomorrow.
Since Friday upto now
- Investigated into the bug related to SIGTTOU. Did some online reading and experimented around, but to no avail. I have an idea why the bug (might be a feature), but I’m not convinced yet.
- Read through Jeff Bonwick’s paper on
The Slab Allocator.
- Read a bunch of blog posts.
- Wrote a draft cover letter (more like the answer to Tell us about yourself?)
- Also started a survey to understand what is an optimal length for a cover letter to receive a positive response (Survey Here) ; Would be great if you could take it. It has 4 questions only.
- Read up on asyncio
- Spend some more time on the SIGTTOU bug
- Maybe work on the slab allocator
- Started reading up about Python internals. Read a blog post on symbol tables.
- Did a rubber duck explanation to
@Dan Luu (W'13)of the shell and the bug with SIGTTOU in the hope of having an “aha” moment. That never came unfortunately. However, we used
straceand we know that SIGTTOU is being emitted for a foreground process when it attempts to write to the terminal. However we noticed a similar behaviour in bash, which verifies that it is not unexpected. But the man pages say SIGTTOU is emitted when a background process attempts to write to the terminal. There is most certainly something that I have missed out. My next step to figure this out is to create a minimal reproducible test case and post on Stackoverflow for help.
- Read about asyncio
- Pair with
@Ethan Raymond (S1'17)on his collaborative editor. We will use asyncio.
- Read about Python’s data model.
- Maybe create a reproducible test case for the shell bug if I feel like it.
- Paired with
@Ethan Raymond (S1'17)to use asyncio for the editor. We hit a dead end, since none of us had worked with asyncio previously. Turns out you can’t do asynchronous input from stdin using asyncio yet. Or the resources that I stumbled upon, weren’t very clear.
- Another dead end: Wanted to integrate the shell into yehos. Realised it doesn’t ship with libc yet, which is a fair amount of work at this point.
- Finalised my second project at RC. I will implement RAFT in Erlang. Both of these are unknown to me, so I will be reading up on RAFT while working through Learn You Some Erlang at the same time. Will be some time before I get started with the actual implementation.
- Went through Jason Eisner’s
How to Read a Technical Paper(https://www.cs.jhu.edu/~jason/advice/how-to-read-a-paper.html) as recommended by
@Naomi Saphra (SP2'17).
- Started reading through the original RAFT paper. I’m taking
extensive notes while reading it, as recommended by
@Naomi Saphra (SP2'17). It surely helps in staying focussed while understanding better instead of just skimming the paper. 33% through the paper at the moment. But the thick of the material starts about now, so it will only take longer as I go deeper into the paper.
- Continue working on the RAFT paper.
- Switch to
Learn You Some Erlangin between.
- Also read up on Python internals
- This looks to be my daily routine, at least for the next few days.
- I will also have to read Leslie Lamport’s paper on vector clocks. Probably won’t start on it today. But surely before I start working on the RAFT implementation.
- RAFT: Worked through the section on log replication.
- Went through the first 5 chapters of Learn You Some Erlang. More like brushing up, since I had done those just before coming to RC.
- Realised that I’ll have to work through most of that book before implementing RAFT.
- Read more of Learn You Some Erlang.
- Let what I’ve learned so far about RAFT and Erlang sink in.
- Read up on Python internals
- Worked through two chapters of LYSE: Higher Order Functions and Errors and Exceptions.
- Woke up pretty late.
- Go see MOMA today at 4.
Friday, Weekend and upto now
- Saw Van Gogh and Picasso’s works at MOMA. Surreal.
- Shell keeps calling me back. Wrote Writing a UNIX Shell - Part III. Took quite some time. Also I learned a few more things about signal handling while writing the blog post itself.
@Saul Pwanson (SP2'17)pointed out a statement from the blog post that turned out to be incorrect. Digged deeper to understand that and updated the post.
- Saw the #NYCPrideParade. Awesome thing. Lucky I got to see it.
- I want to work on the shell more.
- But I also want to do Erlang.
- Toss a coin and decide between the two?
Since last checkin (26th June)
- Paired with
@Kate Murphy (S1'17)on her shell. We understood pipes and when to close the file descriptors. I was also able to validate my understanding of fork().
- Paired with
@Rudi Chen (S1'17)to implement load balancing on his stressed syllables app. Learned a few things about the Elixir which will help me with my own study of Erlang. Elixir definitely has a better syntax.
- Read the internals of Python and gotchas. Specifically the Python data model, variable scoping. I don’t think I understand all of the data model yet. Will probably have to read that a dozen times until I get it.
- Read some other topics: ALB vs ELB. HTTP/2. Docker service discovery. VPC.
Today and the week ahead
- Try to flip the body clock back to saner timings. Currently it starts at 1 in the afternoon and ends at around 5 in the morning. Would like to change this to end by 1 AM instead.
- Get back to Erlang + RAFT. Pick up speed on this.
- Pair some more.
- Wrote a blog post on the RC experience so far.
- Attended my first EDM concert at Central Park.
- Flipped body clock back to regular timings. \m/
- Worked through chapters 8,9 and 10 of Learn You Some Erlang. (Functionally solving problems, A short visit to common data structures, The hitchhiker’s guide to concurrency)
- Terrible dinner at Yiwanmen. (No wonder this was the only empty space while everywhere else was crowded for dinner)
- More Erlang. Didn’t do enough yesterday. Need to do more.
- Revise through my notes on the RAFT paper, upto what I’ve read so far.
- Revised through my notes on RAFT and in the process found a few incorrect assumptions. Also managed to file them in neatly in a more organised way digitally.
- Made some more progress on reading the RAFT paper.
- Gave a mock system design interview to
@Veit Heller (S1'17)and
@Rudi Chen (S1'17).Thanks a lot to both of you for taking out time for this!
- Saw the fireworks on the east side. Pretty amazing.
- Sadly not enough time for Erlang yesterday.
- Made peace with the fact that my estimation skills are terribly out of mark and I end up underestimating the time required for most things all the time. I do want to get better at this, but this is probably one of those things that come with time maybe?
- Set more goals for the rest of the week
- Summer 2’s in the house! yay!
- Participate in the panel on “Choosing and scoping projects”
- Read RAFT (Hopefully finish it?)
- Sat on the panel for “Choosing and Scoping Projects”
- Gave a presentation on the shell I built in my first half of RC
- Surprisingly had a reasonably productive time and was able to finish the RAFT paper. Have a few blind spots in my understanding though.
- Discuss the RAFT paper with
@Rudi Chen (S1'17)
- Pair with
@Rudi Chen (S1'17)on his stressed syllables project. We will work on caching the syllables across servers.
- Squeeze in Erlang in between all of this.
- Paired with
@Rudi Chen (S1'17)to implement caching on the stressed syllables app.
- Paired with
@Elizabeth Rosenzweig (S2'17)to refactor a very messy code base. This is only getting started. It is a lot of work.
- Paired with
@Veit Heller (S1'17)to pick up some ideas on how to better refactor C code. Specifically learned about macros.
- Worked through 2 chapters of Learn You Some Erlang.
- More Erlang.
- More pairing.
- Paired with
@Anna Taberski (S2'17)on understanding a bug in her HTTP server. Learned a few things along the way about golang’s
Connobject. This took a couple of hours somehow. Found a workaround that let’s Anna not get blocked on this. The workaround is hackish but also feels
- The bug managed to spike my curiosity, I did some research after the pairing and suggested her a few more approaches.
- Interestingly, as Anna found, the bug only appears on Chrome and not on Safari.
- Felt tired, probably because of the long week.
- Slogged through a couple of chapters of Learn You Some Erlang and started reading the chapter on concurrent programming from Joe Armstrong’s Programming Erlang as well.
- I feel more or less okay to get started implementing RAFT now. Don’t want to get bogged down to just specialising over Erlang without building anything concrete. Just going through the tutorials gives a feeling of not having accomplished anything.
- Some discussion on the perception of art and literature to end the day.
- Watched Harry Potter and the Deathly Hallows (Part II) and simultaneously echoed the dialogues and waved my arm around when the characters cast spells. :D
- Resumed reading Hackers & Painters. I had stopped midway for some abrupt reason before starting my RC batch. Finishing the book is one of the goals I set before coming to RC.
- Took the Staten Island Ferry and back. Highly recommended. This was something different and much needed as my stay in New York was starting to feel a bit monotonous. Oh, in case you didn’t know, the ferry service runs 24x7 and is free. \m/ ; And it is recommended to take it during the non rush hours.
- Worked through https://rustbyexample.com/ upto section 8. I had always wanted to feel what Rust is all about and I really wanted to work on something completely new. Offers some fresh perspective. I like the syntax of Rust and it also has pattern matching. That makes me excited about trying this out in the future. But for now, it will have to wait.
- More Erlang - specifically read about the Erlang OTP, gen_server and gen_fsm. I have to accept this is all a bit too intense and taking longer than I thought.
- I am close to get started with implementing RAFT (I thought it would be today)
- Spent some time in the evening thinking about my implementation, i.e. what would be the best approach to get started.
- Hopefully, tomorrow we can begin! :airplane_departure:
- Spent most of the day toiling away with gen_fsm in Erlang. Was able to get started with structuring out the skeleton.
- Midway through this, I realised that gen_fsm has been deprecated in favour of gen_statem. Started reading the man pages for gen_statem instead as a result.
- Can’t call this progress, but still better than what I’ve been able to do recently. There’s still a long road ahead before I can start writing code smoothly for this project.
- Had a nice conversation to wrap up the day with
@Kenneth Alexander Durril (S2'17)on “Why are we at RC” and got a few insights into his work/struggles with his project on PDFs so far.
- Keep hitting at the proverbial wall of Erlang until it starts breaking down.
- Rinse. Repeat.
- Paired with
@Jaseem Abid (S1'17). He helped me understand Erlang’s gen_statem that I will need to implement Raft
- Concepts seem clear now but when I attempted to translate it into code, I was still stuck. So that will require more effort.
@Victoria Kirst (S2'17)'s session on CSS. Key takeaway: I won’t be as intimidated the next time I have to fix my blog’s layout.
- Browsed through
@Anna Taberski (S2'17)code on the HTTP server that she wrote. Suggested a few improvements that she could do along with more stuff that she could add (like tests). But I found the code to be well structured and organized.
- Started working on a library to store large files in memcached just for fun; memcached has a max key size of 1MB. Storing files larger than that requires some black magic under the hood, that my library will do.
@Tim Babb (S1'17)session on ray tracers. I am going to start working on one.
- Shared my initial structure of the memcached chunker library with
@Jaseem Abid (S1'17); Made some improvements based on the discussion and there’s more to follow in the coming days.
- Proud owner of a Kinesis Advantage 2 from today onwards. But it feels like I never learned to type. I am giving myself two weeks to be back at my 100% proficiency. Enjoying the keyboard by each passing keystroke. <3
- Paired with
@Kate Ray (S2'17)on how to manager her specific git repo. I provided her with my thoughts on what might be the best approach even if it does not align with git best practices. One of those scenarios where not following the best practice makes life easier instead.
Worked a bit more on the memcached chunker. It has a set_file and get_file api now. It also stores sha1 hash of each chunk and verifies the hash during the get_file operation to ensure that the file was not corrupted since the last time set_file was used. The code is here if anyone is interested: https://github.com/indradhanush/memcached-chunker/tree/feature/improvements
- Have a few more ideas to implement in this: Use hashes as the keys instead of incremental keys. This would mean duplicate files can be stored under the same key. But it would also have to maintain a reference to the number of files currently referring to that key so that the content is only deleted when none of the files refer to the stored piece anymore. Other things include, benchmarking the sha1 hashing to see what are the limits of the system and to find an optimal number of rounds of hashing to be used. I should also benchmark sha256 at least if not sha512 as well. Current codebase is in python2.7; Should port this to python3.6 instead. Will be interesting to understand how the string manipulation changes since in python 3 everything is unicode by default. So my current model of memory calculation will probably not work. Overall this turned out to be an interesting project to pursue, but also something I am reasonably confident about pulling it off. So this becomes my goto project when nothing else seems to be working. Thus pausing work on this for the next few days. But I don’t mind getting back to this if anyone wants to pair on it or is just curious. Key lesson: fallback projects are nice and work amazingly well to boost your confidence.
- Started researching about the ray tracer project. I am going to do this in Python even though it is not the most performant. Reason being learning C++ right now along with implementing ray tracing will be too much to handle for me. Key lesson: Know thy limits.
@Tim Babb (S1'17)for help understanding the ray tracer basics. An intense drill down session followed where he explained the concepts amazingly well. Turns out I had also forgotten some basic math like what happens when you multiply a negative number with a vector. :( But with Tim’s amazing explanation without giving away too much and the right amount of prodding meant I was able to come up with the answer myself: It flips the direction to the opposite side. Also I am glad I asked him for help, because I realised that I was wrong footed right from the beginning and had I started working without asking him, I’d be working on something completely different and not even remotely close to a ray tracer. Key lesson: Ask for help when you need it.
- A bit comfortable on the Kinesis. But still a long way to go.
- Started working on the ray tracer. Wrote some code to determine rays that intersect a sphere.
- The end result was an all white image with black borders.
@Tim Babb (S1'17)helped to figure out the errors in the code.
- While everything looked fine, there was still one last bug in the
form of a missing bracket around a mathematical operation that
@Wesley Aptekar-Cassels (S2'17)helped to figure out.
- Thanks to all this I have something, but far from perfect.
- My mapping of the world co-ordinates vs screen co-ordinates is obviously wrong. Understand and fix this.
- Make more progress on the ray tracer to start accounting for a light source.
- Fixed the coordinate transformation in the ray tracer. I can generate a black and white image of a sphere at varying distances now.
- Changed my data structure to handle RGB. I was only handling single color values so far. And some basic project management to structure the codebase nicely.
- Started reading Designing Data-Intensive Applications. Read through half of chapter 1. I’m taking notes as I read. So this will take longer.
- Started reading What I talk about when I talk about running by Haruki Murakami in the subway. I’m enjoying the read and can resonate with a lot of the author’s feelings about running in general already. This is going to be a nice read!
- Read more of Designing Data-Intensive Applications.
- Understand how a light source affects the image and implement it in the ray tracer.
- Understood the concept of the surface normal and determining the
color of each point using it with a lot of help from
@Tim Babb (S1'17)
- Implemented it and I have a somewhat buggy image of the surface normal.
- RC birthday celebrations
- Finished Chapter 1 of Designing Data-Intensive Applications
- Skimmed through the paper on DynamoDB.
- Start reading Chapter 2 of Designing Data-Intensive Applications
- Find the bug in my surface normal image. I could use some help on this if anyone would be interested.
- Start reading the DynamoDB paper for real.
- Spoke to
@Jinny Cho (S2'17)and helped her understand the differences between DyanmoDB, MongoDB and Postgres. The discussion revolved around trying to find out the best database for her use case.
- Fixed bugs in the ray tracer. The surface normal image looks a bit
saner but is far from correct. I have this .
@Sukanya Aneja (S1'17)helped find a bug where I was taking just the absolute value of the co-ordinates instead of rescaling my range.
@Hang Lu Su (S1'17)understand operating system processes and why cd is not a sys call but a shell builtin.
- Start reading the DyanmoDB paper.
- Attend the ray tracer workshop.
- Maybe pair with someone on the ray tracer. I’ve tried a bunch of things to debug and am virtually out of ideas now.
- Attended the ray tracer workshop. Had a few doubts answered and a revelation afterwards. I’m not sure what changed, but I was using python’s PIL to save the image and switched to scipy.misc.imsave instead. End result I have images that look more like the surface normal image. There are subtle variations as I increase the size of the image. Also changed hardcoded sphere values to depend on the width of the image instead. This makes testing at various sizes trivial. Here’s the result for a 500 * 500 image. Personally I am happy with the result.
- Read upto section 4.2 of the Dynamo paper.
- Had an early start to the day but something else needed my attention. Ended up missing the jobs workshop. :(
- Continue reading the paper.
- Get back to my Erlang project. I feel mentally prepared for this now.
- It has also been a week on the Kinesis, and I feel reasonably comfortable. Definitely more than my 50% of usual proficiency on the keyboard but I get slightly off when it comes to keyboard shortcuts that I had setup previously. It also takes me sometime to adjust to my older muscle memory when I go back to typing on the internal keyboard.
- Discussed about consistent hashing and replication with
@Nikhith Naidu (S2'17)
- Started digging into gen_statem for my Raft implementation once again.
- Not a very productive day overall, but a great dinner with
@Naomi Saphra (SP2'17),
@Stefan Lesser (S1'17)and
@Jaseem Abid (S1'17)where we discussed from Fahrenheit vs Celsius to is it worth blowing a lot of money on an iPhone instead of an high end android for half the money to why does Apple have such an unapproachable image in the public eye.
- Good amount of walking around 42nd street. Saw the UN headquarters from outside.
- More Erlang.
- Commentated movie night.
- Skipped commentate movie night since I felt productive at that time and didn’t want to break the flow.
- Read the design principles of gen_statem. I don’t understand all the things but I am beginning to develop a feel for it.
- Started building a skeleton for the leader election phase
- Filed my notes on RAFT from notebook to a text file.
- Had a Skype call with
@Joseph George Yiasemides (SP1'17). I showed him my in progress skeleton code of the leader election phase. Asked him about my doubts. He shared a few helpful links to understand OTP better. Thanks
@James J. Porter (S'13)for connecting me with him.
- Go through those links.
- Start Grokking Algorithms
- Read through the chapter on Dynamic Programming from Grokking Algorithms. Loved it.
- Started reading the articles that
@Joseph George Yiasemides (SP1'17)shared with me on understanding OTP.
- Continue reading the articles.
- Hans Zimmer concert.
- The wall of Erlang has started to give way.
- Discussed my ideas with
@Jaseem Abid (S1'17)on the leader election phase before getting writing code. We churned out a few obvious design issues that I didn’t quite know the best “erlang” way.
- Paired with him later to implement a very minimal states where the nodes start as “followers” and then one of them becomes a “candidate” and starts an election requesting for votes. The followers oblige and grant votes.
- This is far from perfect. But definitely a start. I feel motivated again. :victory_hand:
- Paired with
@Joseph George Yiasemides (SP1'17)and showed him what I had. I had a few more doubts that had cropped up when I was playing around with the statem further. We debugged the program and found out the best approach to go forward while discussing alternatives.
- The current approach of using state_function mode of callback appears just right. This helps me to move forward. I was fumbling with the alternative of using the handle_event_function mode and it appears to add a generic interface for handling events. Rather explicitly writing code to handle the events that we want is a better approach.
- Continue building the statem.
- Discussion on Career Progression
- RC Presentations
- Job fair
Friday and weekend
- Made some good progress on the leader election. Started cutting down edge cases. This is fun.
- Paired with
@Rudi Chen (S1'17)on more leader election work. Learned a few techniques of building better APIs in Erlang. I can see how thinking in a FP is significantly different than an imperative language.
- Started reading Chapter 2 of Designing Data Intensive Applications.
- Started digging into writing tests in Erlang. Now is a good time to do that in parallel.
- Commentated movie night. Geeked out on philosophy while watching The Matrix.
- Write more code and tests.
- Read a bit.
- Wrote tests for the follower state and fixed bugs along the way.
- Had a few erlang design pattern related questions in mind that I got answered via IRC.
- A quick pairing session with
@Jaseem Abid (S1'17)to do things the Erlang way. I was using pattern matching for my tests and although it made writing them easy, failure messages were painful to parse. We worked around this by wrapping the pattern matching statements around assert statements instead.
- Rinse. Repeat.
- Wrote a whole bunch of tests for the candidate and leader states of the leader election phase. This helped find bugs along the way while I also implemented a few more features.
- After not running my code for a couple of days, I was relying on my tests to get my code to work, I finally ran the leader election for real. I found more bugs as a result that my tests did not. With three nodes I have smooth leader election and reelections. When I have five nodes, I see a weird pattern and a split vote almost all the time.
- Find the source of the annoyance above and fix it. I hope to have ironed out issues in the leader election protocol by EOD.
- Update: Found the weird bug that was causing the cluster to fail with five nodes. It was a hardcoded list of three nodes. :neutral_face: ** That was quick though. I was fearing something much more brutal. :laughing:
- Sorted out leader election. Fixed a couple of bugs and incorrect implementations along with writing tests.
- Did some unrelated reading.
- Took my foot off the pedal for the day. Chilled out.
- I resumed reading the Dynamo paper yesterday and will continue to do that.
- Start playing around with ideas for implementing log replication.
- Read the dynamo paper on the subway (a small part of it)
- Didn’t really work on my project much but went around talking to others
- Spoke to
@Kate Ray (S2'17)about managing her git repo which has a few special requirements. This was a follow up to a discussion we had had a few weeks ago. I told her about the idea of orphan branches, which now appear to be a closer fit as compared to maintaining multiple parallel branches or multiple folders within the same branch.
@Hang Lu Su (S1'17)with python’s mock library. Mock is tricky to understand and I recalled my own painful struggles with trying to understand mock back in the day. But it is powerful at the same time once you understand it.
- Job fair
- A lot of fun and random discussions in Babbage with fellow recursers.
- Continued reading the dynamo paper on the subway.
- Had a fruitful pairing session with
@Sagar Shah (S2'17)on understanding supervisors. Learned about a few tricky spots while pairing that will come in helpful when I have to deal with them myself.
@Jinny Cho (S2'17)get started on her django project while explaining to her the difference between a virtualenv installation and a postgres installation.
@Elizabeth Rosenzweig (S2'17)told me about floobits, which is a collaborative editor with native extensions for emacs, vim, sublime and atom along with a browser interface. I set this up on my system and then tested it out with her. It was a lot of fun to type together but from the comfort of my own editor. I think this will become my preferred mode of pairing remotely after my batch ends.
@Hang Lu Su (S1'17)with understanding python packages and modules which was followed up with a rather lengthy conversation about structuring packages. I need to learn how to get my point across in a more precise manner in a short span of time.
- Showed my code on leader election to
@Veit Heller (S1'17)and explained to him my design considerations.
- Had a Skype call with
@Joseph George Yiasemides (SP1'17)and discussed about tools and libraries used in Erlang for production code. He shared a bunch of resources and well known Erlang projects for reference.
- Got started with log replication. I’ve been writing tests as I go very strictly, which makes the time to build a feature longer but it is totally worth it, since testing this in the shell manually is annoying.
- Paired with
@Jaseem Abid (S1'17)later in the evening and figured out a better way to make tests deterministic. We can test the cluster at different states on intent. Which is a great thing. Learned a bunch from him!
- Continuing from where I left off yesterday.
Last checkin here. Will move to the #alumni-checkins.
- Wrote some code, nothing concrete.
- Lunch with
- Lockpicking workshop.
- Wonderful dinner at the Japanese Kimchi place we’d gone on our first week at RC. Seems only fitting that RC ends with that as well.
- Long conversations late into the night.
- Not write code.
- Feelings checkin.
- End of batch picnic.