IOOPM Project Work
Table of Contents
Note: I am in the process of translating the material for this course from Swedish to English. The reason for this is to be able to have non-Swedish speaking TAs on this course. Rather than delay things or offer only a single project, I have elected to mix Swedish and English where I did not have the time to rewrite old material.
1 Why a Project?
The IOOPM project serves several goals, but mainly, it serves as an introduction to the inherent problems of software development. So far on the course, and in previous courses like PKD, we have focussed on programming, in the sense of "cutting code". Our main problems have been getting things to run, figthing compilers and libraries, chasing double deallocations or memory leaks, or understanding why a particular method call did not get dispatched in a way we intended. These are problems in the smallBy which I do not mean that they are small problems! – now it is time to face problems in the largeAdmittedly, for a quite small large..
The main challenge of software development is not programming – it is understanding specification, communicating with other members of the team, customers and problem owners, dealing with design decisions of the past that turned out not to be as future proof as intended, estimating time, fighting motivational issues, and delivering on time and on budget with acceptable quality, not just once, but consistently. This is, as it turns out, quite a tall order.
While the domain of the projects was picked to serve a learning purpose, the main goal of the project is not to learn more programming, but to put programming to work in a larger context, and deliver in a way that is sustainableSustainable includes that you are supposed to work in a way that is reasonable to expect of you without destroying you if you were to do it for a couple of years..
What this means is that during the project you should be thinking consistently about these things:
- How can we as a team be reasonably sure to deliver what we set out to deliver at a date we committed to?
- How can we load-balance the workload so that no one is overwhelmed?
- How can we be sure that we are delivering the right thing?
- How do we make sure that the thing we are delivering works correctly?
- How do we communicate effectively in the team?
- How do we deal effectively with team members that are demotivated, angry, stressed about things outside of the project, in a way that is sustainable?
- How do we ensure that the team is working smoothly? (E.g., making sure that the team is a team, that people are not working against each others, that no small cabals are formed, etc.)
And more personally about:
- How do I function in a team? Where are my strengths and weaknesses?
- How should I try to adapt the way I work to work better in a team?
- What motivates me? What demotivates me? How do I stay motivated in a project?
- How good am I at estimating how long a task will take? What is the process that produce my estimations?
- What role do I want to play in a team?
Because the goal of the project is to understand and reflect on the software development process, the main requirement for passing and failing is not whether you manage to complete the project or not. What I request is that you all do your best effort to complete the project, in a sustainable way, and without chucking all the principles of good software development just to be able to demo something at the end.
In practise that means that if you are able to hand something in that only implements a subset of the requested functionality, but that is solid enough for others to build on, you are probably not going to fail. However, this should not come as a surprise to your coachThe role of the coach will be described below.! You must efficiently communicate with your coach so that he or she knows what to expect from your final deliverable, or we will expect that you have completed the entire specification.
In the likely event that you think anything about the project specification is vague, know that it is your responsibility to untangle it. This includes details in the specification as well as requirements to pass this part of the course. Also, in either case, you cannot rely on getting answers that satisfy you.
2.1 How to work
This is impossible to describe, but here are some suggestions:
- Work in sprints, ideally more than two
- No coding in the first day and last two days of the project
- Doing a bad job of dividing tasks will hurt the project
- Don't be afraid to revise plans/tasks/alottment/etc.
- Take 5 minutes to sync daily, even of workdays where you only work on other things (like other courses)
- Spend time on "group norming and forming" – have a kick-off, celebrate your wins (and your losses)
- Avoid making this mistake: "I'm not a coder so I better work on something that does not involve programming"
- Don't panic!
- Don't go dark – keep your coach informed about progress
- If you don't understand XWhere X can be stack traversal, pull requests or Scrum, or…, create a task force of 1–2 people to learn about X and educate the others
- When setting goals, try hard to make the goals measurable (how else will you know if you have met them?)
- Risk taking is less risky when combined with time-boxing
- Strong opinions does not make one right
- Don't buy a single person's whole design – share ownership of ideas, decision power, etc.
- Assigning roles and responsibilities is a really good idea, but such things only work if everyone in the team cooperates and e.g. gives the project manager more power
- Don't lie about your abilities
- Don't lie about (lack of) progress
- If you mess up – don't try to cover it up! For example, if you forget to record time measurements – don't fudge them afterwards because you are supposed to. What will you learn from that?
- If you see a team member struggling without speaking up, help that person come clean
- Any process is better than no process
2.2 How Not To Work
Historically on this course, there has been a tendency to focus on optimisations, and following specifications to the letter from the start. This is a really bad idea. As has been regurgitated since day one on this course, don't over-engineer your designs and implementations from the start as you simply do not have enough information or understanding of the problem at that point.
Example: it used to be the case that the project description was phrased in terms of, "implement X, here is how to do it by bit-manipulation to save memory." This was purposely a setup! The right way to go about this is to first do X in the simplest possible way. That allows testing, integration, etc. and benchmarking how much memory would actually be saved using the bit-manipulation implementation! Lots of student groups got stuck on trying to both understand X and the bit-manipulation at the same time, causing the entire project to be delayed or at times even fail because without X, the it was impossible to get anything else workingOne could successfully argue that this is a testament to my failure as a lecturer to hammer in the importance of getting something working before optimising. One could also successfully argue that students follow specifications too slavishly. One could also argue that optimisations have a certain attraction to some people. Etc..
Planning is hard and cannot be learned from a book! If your approach is to work as hard as possible from the start and then harder (or less hard) when you get closer to the deadline, you are not doing it right. Unless you have a plan, you cannot deviate, and you'll have a hard time communicating to us what our expectations should be of your team's final deliverable. Start by making a plan, and see how it goes. Maybe you will learn that you have a tendency to think that things are simpler than they are. Or that you have a tendency to not realise that physical meetings take time, or that it may be hard to get 20 solid hours of coding done in a 40 hour week! Having expectations and refining them are key. Flying blind is not.
This course is not about teaching you the right way to develop software in teams, but about giving you an understanding of the importance of the whole field of software engineering. Later courses will give you better tools and techniques, but unless you reflect on things nowMeaning, your focus is on the finished product, not on the actual project., you'll get less out of those courses later!
3 Project Infrastructure and Constraints
3.1 Time: Lower bound, upper bound
The project may start at December 4th, but not earlier. The project will be handed in no later than 24 hours before the seminar. For example, if your team's seminar is on the 11th of January at 15:00, the strict deadline is January 10th, at 15:00. This is a strict deadline which may not be renegotiated. The time slots for the seminars will be announced on the course web site or Piazza in December.
We expect that you take some time off during Christmas and New Year. The definition of some time is up to you. For example, in 2017, there are exactly four holidays (December 25 and 26, 1st and 5th of January), in addition to weekends. Feel free to handle this in whatever way you want, as long as you get the work done.
In the past, I have had several panic calls from project members who feel that most of the team is AWOL and does not care during the holidays, and feel obligated to work around the clock to move things forward. This causes tension in the team, is bad for quality, and simply serves to perpetuate an already bad situation. As a team, you are responsible for making sure this does not happen. Take responsibility for the well-being of your fellow team members.
You are invited to reach out to me as often as you need during Christmas and New Year. I might be travelling, but working.
3.2 Resources: The Team
Each team will be about 6 people. You may choose to work however you like in this team. You may use pair programming, or not. You may choose to only work together, all at the same time at the University, or completely distributed, or a mix.
Since team members have signed up for the project, this means they have also signed the statement of intent, promising to pull their weight. Any team member that falls back on this promise can be "voted of the Island". Talk to Tobias if the need arises.
3.3 Resources: The Coach
Varje team får en coach tilldelad sig, för att få hjälp och svara på frågor. Dessa anslås på kursens webbsida/Piazza i samband med grupperna. Tidigt under projektets gång bör teamet ha ett möte med coachen. Vid detta första möte skall teamet presentera sin tänkta (partiella?) högnivådesign för coachen, samt sin planering, d.v.s. hur systemet är uppdelat i delsystem, gränssnittet mellan delsystemen, hur delsystemen är fördelade över programmeringspar, och några första grova deadlines.
Teamet ansvarar för att boka ett avstämningsmöte med sin coach någon gång under projektets gång. Vid detta möte skall teamet kort rapportera om hur arbetet fortskrider, om man räknar med att bli klar i tid, eventuella stora problem, etc. Vid behov kan ytterligare möten bokas. Vid problem skall man i första hand kontakta sin coach.
3.4 Failing to Hand In Something Remotely Satisfactory
A team that fails to hand anything in that is remotely satisfactory will be failed. Each team member will be given a new opportunity next year, as part of a different team.
3.5 Failing to Hand in Something Satisfactory
A team that fails to hand anything in that is satisfactory will be given an opportunity to fix their problems. We will negotiate a new deadline, negotiate demands of the final deliverable, and go from there.
Notably, satisfaction is tied to expectations. This means surprising your coach (and other examiners) by handing in something very poor is a harder fail than an expected, equally poor delivery.
4 What to Hand In
The entire hand-in should be contained in the team's GitHub repository. The repository should have the following structure (see below for additional explaination):
. ├── demo/ ├── doc/ │ └── design.md ├── makefile # top-level makefile ├── proj/ │ ├── code_quality_report.md │ ├── deviations.md │ ├── individual_reflection.md │ ├── team_reflection.md │ └── test_report.md ├── src/ └── test/
4.1 The Code
Should be placed in
./src in the repo. You can have whatever
structure under this directory. It should be possible to build the
code using the top-level make file.
4.2 The Tests
Unit tests, integration tests, etc. are placed in
should be possible to run these with
make test and
make memtest with the top-level makefile.
4.3 Integrations with Existing Code
Both projects ask for integration with existing code. Such code
should be placed in
./demo. The top-level makefile should allow
building this code and link it with the code in
4.4 Design Documentation
In the file
./doc/design.md, describe the design of the system at
a high level. You must start with a high-level description using
some graphical notation and accompanying text. The purpose of this
document is to serve as a starting point for someone that wants to
understand the implementation.
This file, together with
./proj/deviations.md (and, naturally the
code) will be used by the examiner to convince him/herself that
you pass Y68.
4.5 Team Retrospective
In a relatively short text write a team retrospective in which you reflect on how the team has worked together. A report template will be provided for this at a later date. Using this template, you will be asked to reflect on the questions listed at the top of this document, but also about:
- The communication between team members and with people outside the team
- The cooperation between team members
- The coordination between team members with respect to technical tasks
- How the programming tasks were divided, distributed, carried out, load-balanced, etc. – if you used pair programming how did that go? Did you use it for everything?
- How did you go about planning? Were your plans mostly followed?
- How did decision making happen?
- What process did you follow, why and how did that go?
- What are they key lessons learned? (Both negative and positive!)
In a separate section of this document, diskutera arbetet med pull requests (PRs) och löpande granskning av kod som går in i ''master''. Vad ledde detta till? Var det endast merarbete eller tjänade det något till? Hur fungerade färgglada diffar på GitHub som stöd för att se vad som ändrats? Hur ofta knuffades en PR tillbaka för förändringar (eller liknande) innan den fick merge:as? Hur många PR gjordes? Låg de länge i snitt? Arbetade alla par lika med PR? Vad fungerade? Vad fungerade inte?
Sammanställ också samtliga gruppmedlemmars tidsloggar och ta fram en gemensam sammanställning av tidsåtgång och fördelning över olika kategorierThe categories should include at least design, documentation, meetings, implementation, and testing. Testing should not include unit testing, but integration testing and regression testing.. Denna skall visualiseras som ett pie chart.
The team retrospective goes into a file in
This file should convince the examiner that you pass Y64. The text on pull requests, together with the actual pull request log on GitHub, should convince the examiner that you pass Y66.
4.6 Individual Reflection
Varje projektmedlem skall skriva en kort självreflektionAbout
1000–1500 characters. Note – all character counts in this
document are to be interpreted as "as outputted by the
wc command". om sin
prestation. Vilka är dina styrkor och svagheter i ett
projektarbete? Vilka egenskaper bör du förstärka och vilka behöver
du bli bättre på? Titta på listan av individuella frågor i början
av detta dokument.
- Hur fungerar du i ett team? Faller du in i ett särskilt mönster, eller tar du återkommande en särskild roll?
- Får du ut något av att jobba i en grupp, eller jobbar du helst ensam?
- Jämför hur mycket du lärt dig under projektarbetet kontra kursens övriga delar, och fundera över hur du bäst lär dig saker.
- Hur har det gått att logga arbetstid?
- "Biggest fail" och "biggest win", dvs. det största problem som du stötte på och det smartaste eller bästa du gjorde.
All individual reflections go into a file in
This file should convince the examiner that each team member passes Y67.
4.7 Test Report
In the file
proj/test_report.md, explain how you are testing
your project on three levels:
- Unit testing
- Integration testing
- Regression testing
And how you have automated these things. Include a graphical overview of the code coverage. List your 6 most nasty bugs by linking to their issue pages on GitHub.
This file, together with the GitHub issue log, should convince the examiner that you pass Y69.
4.8 Code Quality Report
In the file
./proj/code_quality_report.md, your job is to
convince the examiner that the code passes Y65. You can go about
this in any way that you like that is helpful to the reviewer.
Helpful saves time, meaning that blanket statements such as
"just look at all the code" are not good enoughDiscussing
each function individually also would not save time…. The
examiner wants to know to that you think you have written good
code and why you think that the code is good!
Note that Y65 requires consistently good quality meaning that you cannot concentrate on code quality in just a few places, point to them as good examples in the report, and expect to pass.
./proj/deviations.md, describe all deviations from the full
specification. For every feature X that you do not deliver,
explain why you do not deliver it, how the feature could be
integrated in the future in your system, and sketch the high-level
5 Process Requirements
You can use whatever process you like, as long as you pick one and stick to it. Recommendations are to follow some agile methodology like Scrum or Kanban, but I'm fine with you trying to bite off the Rationale Unified Process or doing a bad old waterfall. You don't have time to learn a complicated process, so quickly pick something that you can grasp fast. There is no right choice – whatever you pick, it will form the basis for your more mature understanding later.
5.1 Trello (or similar)
You are required to use Trello to manage tasks – unless you find a better tool. You get to define what better means, and that means in writing in the final project report. Do invite your coach to your Trello board.
5.2 GitHub for Version Control, Issue Tracking and Pull Requests
Under projektet skall ni använda er av GitHub för att versionshantera koden. Ni kommer att få ett privat konto på Github, som ni måste använda. Av uppenbara och fuskrelaterade skäl får koden inte göras publik eller delas med andra utanför teamet (undantaget coachen och kursledningen). Versionshistoriken på Github visar om versionshantering använts på ett vettigt sätt.
Github har ett utmärkt stöd för issue tracking, d.v.s. buggrapporter och diskussioner kring buggar som ni skall använda. Spårbarhet är oerhört viktigt i systemutveckling, så använd en issue tracker/bug tracker, även om alla sitter i samma rum!
All integration av kod med projektets main branch skall ske via så-kallade pull requests, där någon ber om att få sina ändringar merge:ade med projektet. I samband med varje pull request skall någon form av granskning ske. Är koden bra nog för att integreras? Har den test? Har den rimlig dokumentation? Är den förståelig? Har författaren städat koden ordentligt? När koden väl är inne i systemet är den svår att bli av med, så ta tillfället i akt att skapa en projektkultur där endast bra pull requests accepteras.
5.3 Tools for Making Good Code
- You are required to use
valgrindto check for memory errors. You will be asked to run e.g.
make memtestat the final seminar and show the absence of error for a non-trivial program.
- You are required to write unit tests in some existing unit test
framework, like CUnit. You will be asked to run e.g.
make testat the final seminar to show automated tests running. In combination with running
make test, you will also be asked to provide test coverage metrics using e.g. gcov, and reason about good your tests are. Automating this process is a good idea!
- Code formatting tools (not a strict requirement). Please use a
code formatting tool like
astyleor similar to ensure that code formatting is consistent.
6 Nutshell Summaries of the two Projects
Note that the project specifications are subject to change at any moment.
If nothing else is stated, then the projects are required to work on Linux X86 64 bit, Ubuntu 16.04 (which is what the department provides). Note that developing on Windows or macOS and deploying on Linux is very very brittle in C. In those cases, use a virtual Linux environment for testing.
6.1 Project One
Implement a simple system that helps programmers manage memory in
C applications using a reference counting scheme where reference
counts are managed manually through an APIA fancy word for
library, if you will. that you provide. This project involves
(among other things) wrapping
malloc() to install an object
header in each memory allocation, which is hidden from users,
which stores a reference count for that object, and possibly other
The full project description is here.
6.2 Project Two
Implement a system that automatically manages memory using a tracing-based scheme based on Bartlett's compacting conservative collector. In this system, programmers may define their own heaps and when a heap is full, garbage collection will be triggered and memory scanned for garbage. This involves scanning the stack of your application to find all possible pointers into that heap, which requires a bit of low-level trickery.
The full project description is here.