Tag: #programming

2024-05-29

2024-05-25

2024-04-24

2024-04-24

2024-04-18

2024-04-12

Tell LLM why its task is needed and how the output will be used. You’ll get better results, but programmers often don’t do this because it’s anti-modular. 2024-03-30

Often, content is self-labelling. Removing its label can make the content more scannable. Programmers often over-label. 2024-03-20

Including uncaught errors, unhandled promise rejections, image load failures, Content Security Policy violations, and console.error calls. 2024-03-14

Errors in JavaScript cause an ErrorEvent on window. Preventing the default action blocks console output. Resource errors follow the capture-bubble event model. 2024-03-13

Try setting the root font-size to calc(1rem + 0.25vw) instead of using media queries with fixed breakpoints. 2024-03-12

Estimating viewing distance from screen size to set optimal font-size, but other factors like user mobility, long-sightedness, and short-sightedness complicate the ideal font-size calculation. 2024-03-11

“Session replay” tools record user interactions by capturing DOM changes, external resources, and viewport state, using efficient techniques like the MutationObserver API. 2024-03-09

The “golden rule” of code reviews is to approve improvements. Approve pull requests that fix bugs, even if the implementation isn’t ideal. 2023-10-07

Lean is used to define and prove the commutativity of addition on natural numbers. 2023-06-14

Exporting a TensorFlow model, converting it to the TensorFlow.js format, and running it in the browser with interactive input. 2021-03-09

A TensorFlow 2 program that learns the cost per gram of chocolate by training a linear model on example chocolate bar weights and prices. 2021-02-17

A neural network architecture that can output bounding boxes. It uses a fixed set of “anchor points” and predicts offsets and sizes for each anchor. 2021-02-16

Load the model, convert the input to a tensor, preprocess the tensor to match the model’s expected format, run inference with .predict(), and decode the prediction tensor. 2021-02-15

To replace a GIF with an HTML <video> element in 2021, use a long list of attributes disable unwanted browser behaviors. 2021-01-19

A demo of real-time microphone volume measurement using the Web Audio API’s AnalyserNode. 2021-01-18

How to create an NPM addon package from C++ code using the node-gyp tool and a binding.gyp file, with a JavaScript wrapper module for a more idiomatic API. 2020-12-07

I implement homogeneous coordinates in 3D, including rotation, scaling, translation, and projection matrices, and use them to create an animated 3D house scene. 2020-11-29

Homogeneous coordinates enable a matrix to translate points. I show a six-line JS library. 2020-11-28

A 5-line matrix library, and how to derive it without rote memorization. 2020-11-27

JavaScript class syntax is just syntactic sugar over the traditional prototype-based inheritance. Deconstructing a simple class reveals the underlying constructor functions, prototype assignment, and use of super to call the superclass’s constructor. 2020-11-03

JavaScript has two different “prototype” concepts: an own property with the string key “prototype”, and a parent slot. Don’t confuse the two! 2020-11-02

foo.bar, foo.bar(), or foo.bar = baz - what do they mean? A deep dive into prototypical inheritance and getters/setters. 2020-11-01

A playground to explore the various widths and heights of a <video> element, as reported by different browser APIs, when displaying a webcam stream with dynamically adjustable constraints. 2020-10-23

WebGL’s texImage2D function expects pixels in bottom-to-top order, while browsers provide them top-to-bottom. UNPACK_FLIP_Y_WEBGL fixes this behavior. 2020-10-22

Using a varying to pass the texture coordinates to the fragment shader avoids the need to perform divisions in the fragment shader, improving performance. 2020-10-21

Async iterators use the Symbol.asyncIterator method to define the iteration protocol. 2020-10-19

The MediaStream Recording API converts a MediaStream to a Blob of compressed video and audio. A demo where you can record a 5-second clip. 2020-10-17

Babel is a JavaScript compiler that transforms JavaScript code. Here I use its API programmatically. 2020-10-16

require in Electron works like Node.js, allowing you to load modules, but each Renderer process has its own isolated module state. Relative module resolution depends on whether you use loadFile or loadURL. 2020-10-15

An Electron instance is basically a Chromium instance. It just has some Node.js integration. 2020-10-14

A WebRTC-based group chat application that uses Ably’s publish-subscribe system for signaling. Clients connect directly via peer-to-peer data channels, with Ably facilitating the initial connection. 2020-10-13

A method for calculating a bounding circle around a head, using facial landmarks from BlazeFace. Plus a live demo that you can run on your own face. 2020-10-12

A method for calculating a bounding circle around a head, using facial landmarks from BlazeFace. Plus a live demo that you can run on your own face. 2020-10-11

Involves downloading the model manifest file and associated binary weight files, to avoid downloading large models every application launch. 2020-10-07

Creating a “floating head” effect using WebGL and the BodyPix machine learning model to segment the face from a webcam feed. 2020-10-06

Node.js REPL uses the inspector module to safely evaluate expressions as you type, using V8’s evaluation with throwOnSideEffect: true to avoid executing harmful code. 2020-10-05

UMD modules are a way to write JavaScript code that can be used in any module system (e.g. CommonJS, AMD, or as a global variable). 2020-10-04

“Simplified CommonJS wrapping” feature is a hacky attempt to support synchronous require calls by using regexes. It’s gross. 2020-10-03

NPM packages can contain executables, not just Node.js modules. NPM provides features to help distribute and run these executables, such as the bin field in package.json and the npm run-script command. 2020-10-02

Publishing an npm package @jameshfisher/numsyslib with a stringifyRoman function that converts a number to a Roman numeral. 2020-10-01

NPM is Node.js’s package system, allowing installation of arbitrary stuff, not just Node.js modules. Packages are folders, tarballs, URLs, or version identifiers published to the NPM registry. 2020-09-30

ECMAScript modules and CommonJS modules in Node.js are interoperable. Node.js looks at file extensions and package.json fields to determine the module type. ECMAScript modules can only use import, while CommonJS modules can only use require. 2020-09-29

ES modules are just concatenated source files, with some error checking and variable renaming. Live bindings are an illusion - they’re just ordinary variables at the top-level scope. 2020-09-28

The require function in Node.js loads modules, either local files or external packages. I show the algorithm that searches node_modules directories and package.json files. 2020-09-27

Source maps map generated JavaScript to original source files, allowing debugging of the original code in the browser’s dev tools. 2020-09-26

JavaScript modules enable better dependency management and scoping. A guide for developers like me who still use <script> tags everywhere. 2020-09-25

A WebGL shader that uses BodyPix segmentation to set the alpha channel. 2020-09-24

Applying the BodyPix machine learning model to a video stream for real-time person segmentation, with separate render and segmentation loops to optimize performance. 2020-09-23

The CSS resize property allows any element to be resized, not just <textarea>! But there are some gotchas. 2020-09-22

A demo that uses the BlazeFace model from TensorFlow.js to detect faces in a webcam stream, and draws the detected landmarks on the video. 2020-09-21

A step-away background removal technique using WebGL, which compares both chrominance and luminance differences between the webcam feed and a captured background image to create a transparent mask. 2020-09-20

A WebGL fragment shader that implements the Sobel filter, a method for detecting edges in an image. A demo processes your webcam video stream. 2020-08-31

BodyPix is a TensorFlow model for person segmentation. A demo of BodyPix in the browser. 2020-08-16

WebGL alpha-transparency may appear wrong because of premultiplied alpha. Fix it by setting premultipliedAlpha: false when creating the WebGL context. 2020-08-12

A green screen implementation in the browser using WebGL and chroma key. Includes a live demo. 2020-08-11

An green screen implementation in the browser using WebGL and a chroma key shader. Includes a demo that replaces sufficiently green pixels with magenta. 2020-08-10

A green screen implementation in the browser using WebGL and chroma key. Includes a live demo. 2020-08-09

The order of NSWindow levels is defined by the CGWindowLevelKey constants, with lower numerical values being behind higher ones. The NSWindow.Level enum provides a more limited set of semantic level names, with some levels overlapping. 2020-08-03

Use AVCaptureDeviceDiscoverySession instead of the deprecated AVCaptureDevice.devices(for:) to discover video capture devices. 2020-07-31

Calling setFrameAutosaveName on an NSWindow causes its size and position to be saved to user defaults, overriding the contentRect passed to the constructor. To avoid this, do not call setFrameAutosaveName. 2020-07-10

To resolve “the app shows no response upon launch” in App Review, identify and handle potential sources of fatal errors, such as uncaught exceptions, force-unwrapping nil values, unsafe downcasts, and out-of-bounds array access. Display a helpful error message instead of crashing. 2020-04-24

Simulation shows it’s rational to deliberately infect yourself with COVID-19 early on to get treatment, but after healthcare capacity is exceeded, it’s better to avoid infection. Includes interactive parameters and visualizations. 2020-03-14

Using the MediaRecorder API to record a canvas element as a WebM video file, allowing the user to download the recorded video. 2020-03-13

Simulating an epidemic in WebGPU using a direct index into an array of cells, avoiding the issues encountered with the previous WebGL implementation that used a texture. 2020-03-12

Implementing Conway’s Game of Life in WebGPU, a 2D cellular automaton simulation. Uses compute shaders to update the game state and render the cells. 2020-03-03

A WebGPU-powered simulation of the Rule 110 cellular automaton, rendering a dynamic image on a canvas. 2020-03-02

I write an ArrayBuffer to a canvas by interpreting the buffer as pixel data and using ImageData to provide the dimensions. 2020-03-01

Blurring sensitive information in images may not effectively protect privacy, as the original content can often be recovered using deblurring techniques. 2020-02-29

A spatial SIR model for simulating epidemics, implemented using WebGL fragment shaders. Tracks susceptible, infected, and recovered populations in a 2D grid. 2020-02-23

RNA genomes of viruses like COVID-19 are often sequenced as complementary DNA (cDNA) for practical reasons, though the actual genome is made of RNA with uracil instead of thymine. 2020-02-16

An interactive simulation showing how disease characteristics and human behavior can significantly impact the spread of an epidemic. 2020-02-15

Comparing the genome sequences of the Wuhan coronavirus and a bat coronavirus using diff and NCBI BLAST reveals an 89% similarity. 2020-02-09

An algorithm in Haskell using a breadth-first search to find the reachable set of nodes. 2020-01-24

An O(n) algorithm to check if a binary tree is balanced, by passing up the height from recursive calls. 2020-01-23

We can model “infinite tape” using two stacks. Then we can navigate the list in both directions. This lets us bubble-sort the list in O(n^2) time. 2020-01-22

Implement a queue using two stacks, with one stack for the back of the queue and one for the front. Reverse the back stack onto the front stack when dequeuing to maintain the queue order. 2020-01-21

A Haskell solution to the classic Towers of Hanoi problem. 2020-01-20

A stack with a min operation that runs in O(1) time. Store the current minimum alongside each element, or use run-length encoding to compress the stack of minimums. 2020-01-19

The head is the front of the queue, allowing efficient dequeue. The tail is the back of the list. To enqueue efficiently, we keep a reference to the tail. 2020-01-18

The head of the list is the top of the stack. 2020-01-17

Find the middle node of the linked list using the “runner” trick. Reverse the first half, then compare the nodes pairwise from the middle outwards. Restore the list to its original state before returning. 2020-01-16

A trick to find the middle of a linked list by using two pointers, slow and fast, where fast moves twice as fast as slow. 2020-01-15

Reverse a linked list by treating it as a stack, popping elements and pushing them onto a new list. An implementation in C without a separate stack. 2020-01-14

A C function to partition a linked list around a value x, running in optimal time and space. 2020-01-13

To delete a node from the middle of a linked list, copy the value and next pointer of the next node to the current node, then free the next node. 2020-01-12

Use the “runner” technique with two pointers - one leading by k nodes - to find the kth last element of a singly linked list without knowing the list length in advance. 2020-01-11

To remove duplicates from an unsorted linked list with no extra memory, use a nested loop to check each node against all others, removing duplicates as you go. Time complexity is O(n^2). 2020-01-10

An in-place algorithm to rotate a square matrix by 90 degrees. The key is to rotate 4 corresponding points using a formula, applying this to all points in the top-left corner. 2020-01-09

Compresses a string by replacing consecutive repeated characters with the character and its count. The solution requires two passes to first calculate the length of the compressed string, then generate it. 2020-01-08

A C function to replacing spaces with "%20" in-place, by iterating from the end, preserving true string length. 2020-01-07

A C function to determine if one string is a permutation of another, using a character distribution representation for optimal time and space complexity. 2020-01-06

An in-place string reversal function in C, starting by swapping the first and last characters, then iterating towards the middle. 2020-01-05

A recursive algorithm to determine if a string has all unique characters, with an analysis of its time complexity. 2020-01-04

Glitch’s “Show in a new window” feature uses the window.open API to open a new window, and then manipulates the contents of that window by updating its location property when the source code is edited. 2020-01-01

Exploring my 23andMe genome data - a text file of single nucleotide polymorphisms, not a full DNA sequence, requiring a reference genome to interpret. 2019-12-30

CSS percentages have varying meanings depending on the property. They can refer to the containing block’s width/height, the element’s own font size, or other specific measurements. 2019-12-29

A pixel art game developed over Christmas, with code on GitHub and development on Glitch. 2019-12-28

CSS variables, also called custom properties, allow defining and using dynamic values in CSS. They cascade and inherit like other CSS properties, with the var() function used to reference them. 2019-12-21

Simulated annealing is a variation of trial-and-error optimization that generates mutations of the current best guess, and gradually reduces the magnitude of the mutations over time, similar to how metals cool. 2019-05-28

A mathematical model where the neuron is a function that takes binary inputs and produces a binary output based on a threshold. The model can implement basic logic gates like AND and OR, but not more complex functions like XOR. 2019-05-27

Script detects and exposes your local web servers on localhost and your local network. 2019-05-26

Generators in JavaScript can act as both producers and consumers of data. An example Metric generator batches and sends data using the yield keyword. 2019-05-18

JavaScript’s function* defines a generator function, which can be used to generate and consume data. 2019-05-17

Symbols are a new fundamental data type in JavaScript, introduced in 2015. They are used to create unique object property keys that won’t clash with other properties. 2019-05-16

JavaScript iterators are objects that implement a specific protocol, providing a way to iterate over a sequence of values. The for...of loop is syntactic sugar for manually creating and using an iterator. 2019-05-10

Demo of the user’s webcam in a <video> element, using the navigator.mediaDevices.getUserMedia() API. 2019-05-06

A plugin to create tag pages on a Jekyll blog, and a minimal layout for those tag pages. 2019-05-05

Redis Pub/Sub works with master-slave replication, allowing messages published on the master to be replicated to the slave and received by subscribed clients on the slave. 2019-05-04

A new phishing technique that displays a fake URL bar in Chrome for mobile. A key innovation is the “scroll jail” that traps the user in a fake browser. 2019-04-27

I create a native Node.js addon called addon in C++ using node-gyp. The addon exports a hello() function. 2019-04-20

Installing Electron, creating an HTML web page, and writing the main entry point script to load the page. 2019-04-19

I got started with simple web development, but because enamored with increasingly esoteric programming concepts, leading to a “trough of hipster technologies” before returning to more productive work. 2019-03-23

CSS visited link styles are limited for security reasons, as they could reveal a user’s browsing history. Color can be changed, but getComputedStyle will lie about it. 2019-03-08

Bob invents asymmetric cryptography by playing loud white noise to obscure Alice’s message, which he can cancel out but an eavesdropper cannot. This idea, published in 1944 by Walter Koenig Jr., is the forgotten origin of asymmetric crypto. 2019-02-16

Jekyll build breaks due to Bundler 2 incompatibility. Fixed by removing Bundler 1 version from Gemfile.lock. 2019-01-09

Redis Sentinel provides high availability for Redis. We start a Redis master, then three Redis Sentinel instances. They discover each other, then we trigger a failover. 2019-01-08

Use nc (netcat) to create a web server that returns custom HTTP responses. 2018-12-31

I use a custom blogpost command to quickly create new blog post files with the current date and post title. 2018-12-30

I demonstrate animating a cat walk cycle on an HTML canvas using a sprite sheet image and the drawImage method. 2018-12-29

Cookies are client-side storage that get sent with every HTTP request. A cookie is scoped to a domain suffix, a path prefix, and a time range. The API is old and weird. 2018-12-22

A visualization of how frequently different years are referenced in writing over time, showing patterns in how the past and future are viewed. 2018-12-18

The Morris approximate counter is incremented probabilistically. 2018-12-17

The Google Ngram Viewer API is undocumented, but can be accessed by replacing the graph in the URL with json. This returns a JSON response with the frequency of the searched phrase over time. 2018-11-25

How to use the undocumented Strava Route Builder API to get routes between two points, including a JavaScript example. 2018-10-13

Redis sorted sets are implemented using a hash table for efficient key-value lookups and a skip list for efficient score-based ordering. I show a Haskell implementation that uses a forward map for keys and a reverse map for scores. 2018-04-22

An efficient algorithm to round up to the next power of two in C, using bitwise operations and compiler intrinsics for constant-time performance. 2018-03-30

HTTP keep-alive allows reusing the same TCP connection for multiple requests, avoiding the overhead of opening a new connection. HTTP request pipelining sends multiple requests without waiting for their responses. 2018-03-27

A lambda calculus interpreter in Haskell, using de Bruijn indexing and lazy evaluation, with simple integer literals added. 2018-03-15

An x86-64 program that writes to stdout and exits. 2018-03-10

I write a parser for JimScript, an imperative programming language, in Haskell. The parser goes through three stages: tokenization, nesting, and parsing. 2018-03-09

An interpreter for an imperative language in Haskell, featuring integers, operators, global variables, control flow, and I/O primitives. 2018-03-06

nc can create UDP connections. UDP connections are established when the first data packet is sent, and terminated when the server becomes unreachable. ICMP messages notify the client of an unreachable server. 2018-03-04

NAT allows multiple devices in a local network to share a single public IP address. Routers modify IP and port information in network packets, and manage a translation table. 2018-03-02

Linux uses exponential backoff to retry dropped TCP connections, with a configurable retry limit. 2018-02-27

TCP uses sequence numbers to map unordered IP packets to an ordered byte stream. The sequence number field is 32 bits, but wraps around to 0 after reaching the max value. The sender chooses a random “initial sequence number” during the connection handshake. 2018-02-24

tcpdump captures network traffic. Capture and analysis of a TCP connection, including the 3-way handshake and message exchange. 2018-02-23

The magic shell string 3>&2 2>&1 1>&3- swaps stdout and stderr. It does with the dup2 system call to swap file descriptors. 2018-02-22

A C program that writes “hello, world!” to the console using inline assembly instead of standard library functions, demonstrating direct system call invocation. 2018-02-20

Replacing printf with write, and finally directly using the syscall function with inline assembly. 2018-02-19

IP addresses are mapped to MAC addresses using the Address Resolution Protocol (ARP). The OS maintains an ARP cache to store these mappings. 2018-02-11

Subnets divide the IP address space hierarchically using bitstring prefixes. We check if an IP address is in a subnet using C. 2018-02-10

Reverse DNS lookup maps an IP address to a domain name by querying PTR records. The mapping is not always consistent with the forward DNS lookup. 2018-02-09

ping uses ICMP, or “Internet Control Message Protocol”, to send “echo request” packets and receive “echo reply” packets. 2018-02-07

DHCP dynamically assigns IP addresses to hosts, using a client-server protocol over UDP. It involves a sequence of DORA (Discover, Offer, Request, Acknowledge) messages to obtain and configure a network lease. 2018-02-06

nscd, a local DNS resolver within glibc, is non-standard. Instead, use a local DNS server like named or dnscache. 2018-02-05

getaddrinfo ostensibly does DNS lookups. Sounds simple, but it uses more than 100 system calls! Let’s trace the crazy path of address lookup on Linux. 2018-02-03

tcpdump captures and displays network traffic. An example inspecting DNS requests and responses. 2018-02-01

Methods for hashing multiple values, including hash-then-XOR, concat-then-hash, and serialize-then-hash. The only method that’s collision-resistant is serialize-then-hash! 2018-01-09

A stream cipher in Go that generates an infinite keystream from a shared key using SHA-256. 2018-01-01

osquery allows querying a UNIX system using SQL. 2017-12-11

less uses the terminal’s “alternate screen” feature to clear the screen and display the contents of a file, and then restores the previous screen when exiting. 2017-12-04

Generate a master keypair. Encrypting each diary entry with a one-time shared secret. Encrypt the shared secret with the public key. Decrypt entries using the private key and the shared secret. 2017-12-03

OpenSSL’s inadequate documentation, confusing key formats, and deprecated interfaces make it difficult to use, despite its importance. 2017-12-02

ASN.1 is a data format used to encode structured data like RSA private keys and certificate signing requests. Using openssl asn1parse to reveal its structure. 2017-11-30

The used metric in free -m includes memory used for caching, not just application memory. The true memory usage is lower, calculated as used minus buffers/cache. 2017-11-29

GitHub-Flavored Markdown supports “diff views” with + and - prefixes. While it loses syntax highlighting, it’s a quick and easy way to show changes. 2017-11-27

DNS resolution is a recursive procedure involving different record types like A, CNAME, and NS. Resolving a domain name can require multiple DNS queries to different nameservers. 2017-11-25

Composer’s autoloader uses PSR-4 to map namespaces to directories, allowing you to load classes without explicit require statements. 2017-11-09

Implementing a Promise API in JavaScript, including handling asynchronous callbacks, promise chaining, and automatically wrapping non-promise values. 2017-11-07

To release a PHP Composer package, create a GitHub repo, add a composer.json file, submit the package to Packagist.org, and tag a stable version. Then Composer users can require your package. 2017-11-06

Promises in JavaScript represent future values. They provide a way to handle asynchronous operations, allowing chaining of callbacks. Promises can be fulfilled or rejected, with callbacks handling each case. 2017-11-05

Generating ECDH keys and deriving a shared AES key. 2017-11-03

Generating a key, encrypting and decrypting text, and explaining the implementation. 2017-11-02

A web page that allows you to sign a string using the HMAC algorithm with SHA-256, implemented using the Web Crypto API. 2017-10-31

The Web Cryptography API provides low-level crypto primitives in JavaScript, including hashing strings using SHA-256. 2017-10-30

A WebGL shader that combines diffuse and specular lighting, allowing the user to adjust the relative contribution of each using keyboard controls. Guest post by Luís Fonseca. 2017-10-28

A “hash ratchet” ensures that a compromise of the current key does not allow decryption of past messages. Each message’s encryption key is derived from the previous key using a hash function. 2017-10-27

Businesses often tout “free” services, but the real costs come in terms of time, thought, and typing required from users. Reducing these “Three Ts” is key to improving sign-up flows and increasing conversions. 2017-10-26

Group chat with end-to-end encryption can use “client-side fan-out” where the sender encrypts the message separately for each recipient, or “server-side fan-out” where the sender encrypts the message once using a shared secret, then the server distributes the ciphertext to the group. 2017-10-25

Differences between the Web Push API in Google Chrome and Firefox. Firefox uses the Mozilla Push Service, enabling push notifications without Google’s GCM/FCM. 2017-10-24

Implements a 1024x1024 grid of Conway’s Game of Life using WebGL fragment shaders. 2017-10-23

A Conway’s Game of Life simulation implemented using WebGL fragment shaders, with rendering to texture and a Gosper Glider Gun initial state. 2017-10-22

Rendering an animated 3D cube in WebGL using a vertex shader, face color uniforms, and matrix transformations for rotating the cube. 2017-10-21

Generating dynamic normal-mapped ripple effects using WebGL, with the mouse position controlling the light source. 2017-10-20

GLSL varying allows vertex shader outputs to be passed to fragment shaders. A WebGL demo with colors interpolated between vertices. 2017-10-19

extern declares a variable without defining it, allowing the linker to find the definition elsewhere. This is useful when a variable is declared in one file but defined in another. 2017-08-27

Sentence-wrap is a text-wrapping approach that inserts newlines between sentences. Diffs are easier to read, and sentences more apparent. Reminiscent of poetry. 2017-08-23

#include in C does not import functionality, but rather inserts the preprocessed text of the included file. The actual implementation of the included functions, like fprintf, is provided through linking. 2017-08-21

C’s _Generic is not true generics. It requires manual implementation of concrete type cases, unlike generics that automatically generate code for any type. 2017-08-19

Using the Go "net" package to look up IP addresses for a given domain name, using either the C stdlib or a pure Go DNS resolver. 2017-08-03

Create message digests using the openssl dgst command, specifying the hash algorithm (e.g. -sha512) and optionally signing with a shared password using -hmac. 2017-03-13

The openssl passwd command hashes passwords using the outdated crypt algorithm, with truncation to 8 characters - a poor choice for secure password hashing. 2017-03-12

Use the openssl s_client command to fetch a server’s SSL certificate chain, including the root certificate. 2017-03-11

Generate random bytes with openssl rand, which uses a PRNG seeded with entropy from ~/.rnd. 2017-03-10

Examines Redis’s source code to understand how it tracks subscriptions, handles disconnections, and implements pattern subscriptions, highlighting potential pitfalls and attack vectors. 2017-03-01

Install ebe assembler with a one-liner, but be wary of the SourceForge installation script. 2017-03-01

A TCP server that uses pthread to serve multiple clients concurrently, with an “echo” server for each connection. 2017-02-28

UTF-8 is a character encoding that can represent the entire Unicode character set. It’s variable-length, self-synchronizing, and an extension of ASCII. 2017-02-26

A TCP server that uses the fork system call to create a new child process for each accepted connection, allowing it to handle multiple clients concurrently. 2017-02-25

mode_t is a bitfield type in C that represents file permissions, including read, write, and execute permissions for the owner, group, and other classes. It provides symbolic constants to set and check these permissions. 2017-02-24

A program in C that prints the individual bits of various data types, showing how they are represented in memory. 2017-02-23

ssize_t is a signed version of size_t, used to represent return values from system calls that may need to indicate error with -1. 2017-02-22

A FIFO is a special file that allows inter-process communication. The mkfifo system call creates a FIFO, enabling processes to read from and write to it. 2017-02-21

A “Hello, world!” program in x86-64 assembly for macOS, using the NASM assembler and system calls for writing to stdout and exiting. 2017-02-20

lsof lists open system resources, including pipes, sockets, and yes, files. It shows their type, owner, and location. 2017-02-20

Generate Intel and AT&T assembly with clang, demonstrating the difference in syntax between the two styles. 2017-02-19

setjmp and longjmp in C bypass normal function call and return flow, allowing a function to return to a previously saved state. setjmp saves the current execution context, and longjmp restores it. 2017-02-18

A C function to create a new process, set up its standard input/output/error pipes, and return a struct containing the process ID and pipe file descriptors. 2017-02-17

A talk given with Will Sewell. Overview of Go’s concurrent garbage collector. 2017-02-15

Use the dup system call to duplicate a file descriptor in C, allowing two references to the same underlying pipe. 2017-02-15

UNIX represents various system components, like processes, file descriptors, and memory, using in-memory data structures that can be modeled as relational database tables. 2017-02-14

Lamport timestamps measure time and causality in distributed systems. Timestamps are assigned to events, with the property that if event A happened-before event B, then A’s timestamp is less than B’s. 2017-02-12

The happened-before relation models time in distributed systems, where events at different locations may not have a total order. It defines time using causality. 2017-02-10

To call a program from C, use `fork` then `execve`. There is no more direct way! 2017-02-07

fork duplicates the current process. It returns 0 in the child process. In the parent process, it returns the child’s new process id. 2017-02-06

execve replaces the current process with a new one. It takes a path, an argument array, and an environment array. The process never returns unless execve fails. 2017-02-05

Realtime apps use protocols like HTTP streaming, long polling, WebSocket to enable live updates. Scaling pub/sub requires distributed servers and routing clients to the closest server. 2017-02-04

Compiling with the -S flag generates assembly. 2017-02-03

Access environment variables in C by using the global environ variable, which points to an array of string pointers representing the environment. 2017-02-02

The man command is used to access the UNIX manual pages, which are organized into numbered sections. man searches a predefined path to find the relevant manual page. 2017-01-30

Processes can communicate via files, pipes, signals, sockets, message queues, semaphores, and shared memory. 2017-01-29

To write to a file, open the file, mmap the file descriptor, then write to memory. 2017-01-28

mmap is a system call in C that allows programs to manipulate underlying hardware resources, such as physical memory or files, using ordinary memory manipulation. 2017-01-26

How to check if a 64-bit word contains a zero byte. A step-by-step example and a proof of correctness. 2017-01-24

To subtract in binary, we use carry bits and two’s complement representation. 2017-01-24

Integers in C are int by default, while suffixes like L and U denote long and unsigned types. Floats are double without a suffix. 2017-01-23

Literals are lvalues with addresses, while constants are rvalues without addresses. In C, only string literals are literals; other “literals” like numbers and characters are constants. 2017-01-22

“lvalue” either means “expression which can be placed on the left-hand side of the assignment operator”, or means “expression which has a memory address”. “rvalue” is defined as “all other expressions”. 2017-01-21

The UINT64_C macro appends the ULL suffix to a 64-bit unsigned integer literal, converting it to the appropriate type. 2017-01-20

The RTCDataChannel API lets us configure the delivery guarantees, including the ordered, maxPacketLifeTime, and maxRetransmits properties. 2017-01-17

Including setting up the RTCPeerConnection, creating a data channel, handling ICE candidates, and generating an offer to be shared with the remote peer. The signaling channel is copy-paste! 2017-01-16

C signal handlers reuse the same call stack as normal functions, but can optionally use a special signal stack instead. 2017-01-14

sigaction says what to do when a signal is received. signal is a simplified interface to sigaction. 2017-01-13

Executing do_something() N times in C using a while loop and post-decrement operator. 2017-01-12

To unregister a signal handler in C, use signal(signum, SIG_DFL) to reset the disposition to the default, or signal(signum, SIG_IGN) to set the disposition to ignore the signal. 2017-01-11

signal returns the old signal handler pointer for given signal number. 2017-01-10

Signals in C are software interrupts that allow programs to respond to events. The signal.h header provides functions to register signal handlers and send signals. 2017-01-09

Using unique error codes with linkable documentation improves error reporting. Provide a central error page URL with each error message. 2017-01-05

Bitfields in C allow compact bit packing in structs, avoiding manual bitwise operations for greater safety and clarity, at the cost of some language complexity. 2017-01-04

A union in C allows one variable to hold different data types, where the storage is shared. The size of a union is at least the size of its largest member. Fields in a union have the same byte offset. 2017-01-03

Efficient packing of game player data into 16 bits using bitwise operations on a uint16_t type. 2017-01-02

GCC’s ‘statement expressions’ allow inserting statements into expression positions. Behavior is unspecified. 2016-12-30

Redis uses length-prefixed strings with pointers into the middle of the allocation, allowing C-string operations on the string data. 2016-12-28

C allows a single array field of incomplete type (char []) at the end of a struct definition. The size of the struct is determined by the size of the fixed fields, and the dynamic-sized array is allocated separately. 2016-12-27

Use the times() system call in C to measure the CPU time used by a process, distinguishing between time charged to the process itself and time charged to the kernel on its behalf. 2016-12-26

C array literals can use explicit indexes. The array length is determined by the largest explicit index. 2016-12-25

perror in C prints a human-readable description of the last error that occurred, as stored in the global errno variable. 2016-12-24

Useful to show how C represents various data types. 2016-12-22

A C program that creates an HTTP server that responds with “Hello, world!” to every request. 2016-12-20

The syscalls needed for a simple UDP echo server are socket, bind, recvfrom, sendto, and close. 2016-12-19

Kqueue is a more efficient alternative to select for managing multiple TCP connections, providing a publish-subscribe model for tracking events in the kernel. 2016-12-18

fdset in C is a data structure that represents a set of file descriptors. The FD_SET, FD_CLR, FD_ISSET, and FD_ZERO macros are used to manipulate and inspect these sets. 2016-12-17

The select syscall allows a process to sleep and wake up when a file descriptor is ready for reading, writing, or has an exceptional condition. This enables a TCP server to handle multiple clients concurrently. 2016-12-16

File descriptors should be called “resource descriptors”. They represent various system resources, not just regular files. The operations you can perform depend on the specific resource. 2016-12-15

A minimal TCP server in C uses the socket, bind, listen, accept, recv, send, and close syscalls to manage connections. 2016-12-14

errno lets you access error codes from system calls. It’s a global int. 2016-12-13

static functions in C are only callable within the translation unit they are defined in, not across the whole program. 2016-12-12

Use i & (n-1) to compute i % n, when n is a power of two. This avoids the modulo operation. 2016-12-10

Macros in C can define token replacements or function-like replacements. 2016-12-09

Arrays in C can “decay” to pointers, but are not inherently pointers. The size of an array is lost during decay. 2016-12-08

Automatic variables in a Makefile, like $@ for the target and $^ for the dependencies, reduce repetition in rules. 2016-12-07

C strings use null-terminated byte arrays, which cannot represent arbitrary binary data. Binary-safe strings explicitly store the length, allowing them to represent any sequence of bytes. 2016-12-06

Set the C compiler in Makefile using predefined variable $(CC), which defaults to cc and can be redefined. 2016-12-05

A FILE is a C data type that represents an open file. It contains metadata about the file, such as its file descriptor and buffers. 2016-12-04

The restrict keyword in C is a type qualifier that indicates a pointer parameter points to an object not accessed through any other pointer in the function’s scope. This allows the compiler to make optimizations. 2016-12-03

const is a type qualifier in C that makes a variable unassignable, except during initialization. 2016-12-02

C99 introduced the stdbool.h header, providing bool, true, and false for boolean values, building on the _Bool type. 2016-12-01

realloc resizes an allocated memory block, copying contents if necessary. 2016-12-01

C is defined in several places, including “The C Programming Language” by Kernighan and Ritchie, and a series of ISO C standards (C89, C99, C11). 2016-11-30

Computing a pointer to unowned memory invokes undefined behavior, even without dereferencing! 2016-11-30

Use concatenated string literals with newline characters to define multi-line string literals in C. 2016-11-30

C string literals can contain comment-like characters, which are treated as literal text, not as comments. 2016-11-30

C array subscripts are defined as pointer arithmetic, where a[b] means *(a+b). This means a[b] is the same as b[a]! 2016-11-30

size_t is used to represent the size of objects in memory. It is commonly used for array indexing. 2016-11-29

Use uintptr_t to count objects in C, as it can represent the largest possible pointer value. 2016-11-29

static in C can modify variable declarations inside and outside a function body, and function parameters. 2016-11-28

void in a C function parameter list means the function takes no arguments, whereas an empty list allows for an unspecified number of arguments. 2016-11-27

void in C has multiple meanings: as a return type to indicate no return value, as a parameter type to indicate no parameters, and as a pointer type to indicate a pointer to an unknown type. 2016-11-27

The “K&R-style” function definition in C uses an initializer list to declare parameters, unlike the more common parameter type list. This old-style definition has significant semantic differences and should be avoided due to its potential for undefined behavior. 2016-11-27

A convention for expressing complex C types using typedef and a “reverse Polish notation” syntax to improve readability. 2016-11-24

The stack layout in C includes function arguments, return address, and local variables. Addresses decrease as you go through the arguments, and the stack grows downward with each function call. 2016-11-24

Variadic functions in C use the ... syntax to accept an arbitrary number of arguments. The stdarg.h library provides va_start, va_arg, and va_end to access the arguments. 2016-11-23

An .xcworkspace file is a directory that contains a workspace configuration for Xcode projects. 2016-11-17

The .xcodeproj “file” is actually a directory that contains a project.pbxproj file, which is a text file that stores the settings of an Xcode project, including information about the files, build configurations, targets, and dependencies. 2016-11-17

I try Vim. I give up after discovering that the cursor cannot sit at the end of a line. 2016-11-16

Serialize JSON in Swift with JSONSerialization.data(withJSONObject:options:). 2016-11-16

Golang’s garbage collector uses a “tricolor” algorithm, dividing heap objects into black, white, and grey sets. The algorithm can run concurrently with the program, and the “tricolor” invariant ensures no pointers go directly from the black set to the white set, allowing the white set to be cleared. 2016-11-11

Large working sets and low latency are incompatible with GHC’s stop-the-world garbage collector, which optimizes for throughput instead of latency. The collector’s pause times scale linearly with working set size. 2016-05-12

2015-01-26

2014-10-25

2014-10-11

Nix is a functional programming language used for package management. We see the Nix expression language, including strings, primitives, operators, functions, let expressions, and conditionals. 2014-09-28

2014-09-19

2014-09-13

Prolog as a better alternative to “semantic triples” and SPARQL. 2014-08-24

2014-08-19

2014-08-15

Documentation often ends up in black holes - internal documents, issue trackers, test cases, conversations, commit messages, and wikis - where it goes unread. Focus on writing documentation that people will actually read, like product names, homepages, and inline help. 2014-08-13

2014-08-10

2014-07-27

Syntax highlighters make value judgments about code. Most highlighters judge that comments are cruft, and try to hide them. Most diff viewers judge that code deletions are bad. 2014-05-11

A JS function to decide the Halting Problem will always have a logical contradiction. 2013-12-24

A visual programming language inspired by Haskell. 2012-02-27

2010-01-19