SymPy Updated 2025-07-16
This is the dream cheating software every student should know about.
It also has serious applications obviously. www.sympy.org/scipy-2017-codegen-tutorial/ mentions code generation capabilities, which sounds super cool!
The code in this section was tested on sympy==1.8 and Python 3.9.5.
Let's start with some basics. fractions:
from sympy import *
sympify(2)/3 + sympify(1)/2
outputs:
7/6
Note that this is an exact value, it does not get converted to floating-point numbers where precision could be lost!
We can also do everything with symbols:
from sympy import *
x, y = symbols('x y')
expr = x/3 + y/2
print(expr)
outputs:
x/3 + y/2
We can now evaluate that expression object at any time:
expr.subs({x: 1, y: 2})
outputs:
4/3
How about a square root?
x = sqrt(2)
print(x)
outputs:
sqrt(2)
so we understand that the value was kept without simplification. And of course:
sqrt(2)**2
outputs 2. Also:
sqrt(-1)
outputs:
I
I is the imaginary unit. We can use that symbol directly as well, e.g.:
I*I
gives:
-1
Let's do some trigonometry:
cos(pi)
gives:
-1
and:
cos(pi/4)
gives:
sqrt(2)/2
The exponential also works:
exp(I*pi)
gives;
-1
Now for some calculus. To find the derivative of the natural logarithm:
from sympy import *
x = symbols('x')
print(diff(ln(x), x))
outputs:
1/x
Just read that. One over x. Beauty. And now for some integration:
print(integrate(1/x, x))
outputs:
log(x)
OK.
Let's do some more. Let's solve a simple differential equation:
y''(t) - 2y'(t) + y(t) = sin(t)
Doing:
from sympy import *
x = symbols('x')
f, g = symbols('f g', cls=Function)
diffeq = Eq(f(x).diff(x, x) - 2*f(x).diff(x) + f(x), sin(x)**4)
print(dsolve(diffeq, f(x)))
outputs:
Eq(f(x), (C1 + C2*x)*exp(x) + cos(x)/2)
which means:
To be fair though, it can't do anything crazy, it likely just goes over known patterns that it has solvers for, e.g. if we change it to:
diffeq = Eq(f(x).diff(x, x)**2 + f(x), 0)
it just blows up:
NotImplementedError: solve: Cannot solve f(x) + Derivative(f(x), (x, 2))**2
Sad.
Let's try some polynomial equations:
from sympy import *
x, a, b, c = symbols('x a b c d e f')
eq = Eq(a*x**2 + b*x + c, 0)
sol = solveset(eq, x)
print(sol)
which outputs:
FiniteSet(-b/(2*a) - sqrt(-4*a*c + b**2)/(2*a), -b/(2*a) + sqrt(-4*a*c + b**2)/(2*a))
which is a not amazingly nice version of the quadratic formula. Let's evaluate with some specific constants after the fact:
sol.subs({a: 1, b: 2, c: 3})
which outputs
FiniteSet(-1 + sqrt(2)*I, -1 - sqrt(2)*I)
Let's see if it handles the quartic equation:
x, a, b, c, d, e, f = symbols('x a b c d e f')
eq = Eq(e*x**4 + d*x**3 + c*x**2 + b*x + a, 0)
solveset(eq, x)
Something comes out. It takes up the entire terminal. Naughty. And now let's try to mess with it:
x, a, b, c, d, e, f = symbols('x a b c d e f')
eq = Eq(f*x**5 + e*x**4 + d*x**3 + c*x**2 + b*x + a, 0)
solveset(eq, x)
and this time it spits out something more magic:
ConditionSet(x, Eq(a + b*x + c*x**2 + d*x**3 + e*x**4 + f*x**5, 0), Complexes)
Oh well.
Let's try some linear algebra.
m = Matrix([[1, 2], [3, 4]])
Let's invert it:
m**-1
outputs:
Matrix([
[ -2,    1],
[3/2, -1/2]])
Hadron Updated 2025-07-16
Big O notation family Updated 2025-07-16
This is a family of notations related to the big O notation. A good mnemonic summary of all notations would be:
LeetCode Updated 2025-07-16
Their system is quite good actually. Not as good as a GitHub repo with all the tests made explicit. But still pretty good.
Text-based user interface Updated 2025-07-16
The perfect Middle Way between command-line interfaces and GUIs. A thing of great beauty.
Cloud computing market share Updated 2025-07-16
Figure 1.
Cloud Computing market share in Q2 2022 by statista.com
. Source.
Hyperscale computing Updated 2025-07-16
Basically means "company with huge server farms, and which usually rents them out like Amazon AWS or Google Cloud Platform
Figure 1.
Global electricity use by data center type: 2010 vs 2018
. Source. The growth of hyperscaler cloud vs smaller cloud and private deployments was incredible in that period!
Platform as a service Updated 2025-07-16
Highly managed, you don't even see the Docker images, only some higher level JSON configuration file.
These setups are really convenient and cheap, and form a decent way to try out a new website with simple requirements.
Closed standard Updated 2025-07-16
ISO is the main culprit of this bullshit, some notable examples related to open source software:
The only low level thing that escaped this was OpenGL via Khronos, what heroes those people are.
How the hell are you supposed to develop an open source implementation of something that has a closed standard?
Not to mention open source test suites, that would be way too much to ask for, those always end up being made by some shady small companies that go bankrupt from time to time, see e.g. .
Inner source Updated 2025-07-16
If you are going to do closed source, at least do it like this.
Basically the opposite of need to know for software.
Closed source on offline products used by millions of people is evil, when you could just have those for free with open source software! Thus Ciro's hatred for Microsoft Windows and MacOS (at least userland, maybe).
Some anecdotes.
Ciro Santilli never splits up functions unless there is more than one calling point. If you split early, the chances that the interface will be wrong are huge, and a much larger refactoring follows.
If you just want to separate variables, just use a scope e.g.:
int cross_block_var;

// First step.
{
    int myvar;
}

// Second step.
{
    int myvar;
}
Ciro has seen and had to deal with in his lifetime with two projects that had like 3 to 10 git separate Git repositories, all created and maintained by the same small group of developers of the same organization, even though one could not build without the other. Keeping everything in sync was Hell! Why not just have three directories inside a single repository with a single source of truth?
Another important case: Linux should have at least a C standard library, init system, and shell in-tree, like BSD Operating Systems, as mentioned at: Section "Linux".
A slow development test cycle will kill your software.
New developers won't want to learn your project, because they would rather shoot themselves.
This means that build time, and the time to run tests, must be short.
5 seconds to rebuild is the maximum upper limit.
Of course, at some point software gets large enough that things won't fit anymore in 5 seconds. But then you must have either some kind of build caching, or options to do partial builds/tests that will bring things down to that 5 second mark.
You also have to spend some time profiling execution and build from scratch times.
A slow build from scratch will mean that your continuous integration costs a lot, money that could be invested in a new developer!
It also means that people won't bother to reproduce bugs on given commits, or bisect stuff.
One anecdote comes to mind. Ciro Santilli was trying to debug something, and more experience colleague came over.
To reproduce a problem, ciro was running one command, wait 5 seconds, run a second command, wait 5 seconds, run a third command:
cmd1
# wait 5 seconds
cmd2
# wait 5 seconds
cmd3
The first thing the colleague said: join those three commands into one:
cmd1;cmd2;cmd3
And so, Ciro was enlightened.
Figure 1.
xkcd 303: Compiling
. Source. They should be benchmarking and fixing their shitty build system instead.

There are unlisted articles, also show them or only show them.