Explain This Code

Recursion. factorial(n) returns 1 when n <= 1 and otherwise returns n * factorial(n - 1). Calling factorial(5) starts a chain of calls: factorial(5), factorial(4), factorial(3), factorial(2) and factorial(1), which is 5 calls. The check n <= 1 runs once per call, so 5 times: false for 5, 4, 3 and 2, true only for 1, the base case that stops the chain. The answers then come back up: 1, 2 × 1 = 2, 3 × 2 = 6, 4 × 6 = 24 and 5 × 24 = 120. The tool reports exactly these numbers from its run (5 calls; true once, false 4 times; prints 120 after 24 recorded steps).

A closure. makeCounter() creates count = 0 and returns an inner function that adds 1 to count and gives it back. const next = makeCounter(); keeps that inner function, and with it access to its own count, even though makeCounter has finished. The two bare next(); calls give back 1 and then 2, which are thrown away because nothing stores or prints them; console.log(next()) makes count 3 and prints 3. Reading it as “count is reset to 0 on every call” would predict 1, which is why this snippet confuses people.

The common mistake: an off-by-one loop. With const arr = [5, 10, 15]; the loop for (let i = 0; i <= arr.length; i++) runs while i is 0, 1, 2 and 3: 4 passes for 3 items. The last pass reads arr[3], which does not exist and is undefined, so sum goes 5, 15, 30 and then 30 + undefined = NaN. The tool flags the <= with .length as one past the last item and shows the printed NaN; with i < arr.length the loop makes 3 passes and prints 30. The C preset is not run here, but it is easy to check by hand: 4 + 8 + 15 + 16 + 23 = 66, so printf("Sum: %d\n", sum) would print Sum: 66.

Two kinds of explanation. JavaScript and Python inside the Code Lab’s teaching subset are parsed, so every sentence comes from the exact structure of the code: which variable a loop counts with and between which bounds, which branch an else belongs to, whether a function calls itself or returns another function. The same engine then runs the program, so explanations quote the real output (“When run, this prints 120”), how many passes each loop made and how often each check was true. Code outside the subset, such as classes, list comprehensions or f-strings, falls back to line patterns and the page says so. Everything else is explained by line patterns per language and is labelled pattern-based — may miss details.

Where the language facts come from. JavaScript: MDN Web Docs (closures, map, filter, reduce, sort, which compares items as strings unless given a compare function, slice, indexOf, String.replace, which replaces only the first match of a string pattern, let, const, strict and loose equality, hoisting and Math.random). Python: the Python tutorial and library reference (range never includes its end point, list comprehensions, / always returns a float and // floors, negative indexes count from the right, docstrings, str.replace replaces every occurrence, split() with no separator splits on runs of whitespace, dict.get defaults to None, list.sort sorts in place, round() rounds halves to even). C: cppreference for the printf conversions (%d, %f with 6 decimals by default, %c, %s), the main function and its zero exit status, and switch fall-through. SQL: the PostgreSQL documentation for SELECT processing, joins, count(*), LIKE wildcards, BETWEEN (inclusive), PRIMARY KEY, and UPDATE, DELETE and DROP TABLE. Java’s main method follows the Oracle Java Tutorials, C#’s Main the Microsoft C# documentation, PHP’s foreach the PHP Manual, and shell syntax, test operators, pipelines and command substitution the GNU Bash manual, with the GNU Coreutils manual for ls, cp, mv, rm, mkdir, cat, head, tail, sort, uniq, wc, chmod and touch.

What it cannot tell. A pattern sees one line at a time, so it does not know what a variable holds, which overload a method call picks, or what a macro expands to. It never runs C, Java, SQL or shell commands, and the JavaScript and Python stepper runs a teaching subset, so a program that relies on something outside it is explained without its output. The explanation describes what the code does, not whether that is what you meant: an off-by-one is only flagged in the common <= .length form.

The idea that a program should explain itself is old. The preface to the first edition of Abelson and Sussman’s Structure and Interpretation of Computer Programs (MIT Press) puts it as a principle: “programs must be written for people to read, and only incidentally for machines to execute.” In 1984 Donald E. Knuth published “Literate Programming” in The Computer Journal 27(2), pages 97–111 (doi:10.1093/comjnl/27.2.97); his description of the later book of the same name says the main idea is “to treat a program as a piece of literature, addressed to human beings rather than to a computer.”

Reading code is also a skill that computing education studies on its own, separately from writing it. In 2004 an ITiCSE working group of 12 authors led by Raymond Lister published “A multi-national study of reading and tracing skills in novice programmers” (ITiCSE working group reports, pages 119–150, doi:10.1145/1044550.1041673). Tracing what a program does step by step is what Philip Guo’s Online Python Tutor automated (Proceedings of the 44th ACM Technical Symposium on Computer Science Education, 2013, pages 579–584, doi:10.1145/2445196.2445368); the Code Lab’s Execution Visualizer does the same in your browser, and this page adds the sentence-by-sentence reading on top of the same engine.

Reference: Marlin G-code Reference.

Reference: M-Code Reference by Controller.

Explain This Code turns a snippet into plain English: a one-paragraph summary of what the whole program does, then an explanation of every statement, indented to match the code, with notes on the operators and built-in functions it uses. For JavaScript and Python it also runs the code and quotes the real output, the number of loop passes and how often each condition was true. It reads 16 languages and explains C, C++, Java, C#, PHP, SQL, Bash, HTML, CSS and JSON from recognised patterns.

Most code explainers produce fluent text that can be confidently wrong. This one only says what a rule or the syntax tree supports, marks anything it does not recognise, labels pattern-based explanations as such, and shows the rule behind every row in Show Work. It also connects to the rest of the Code Lab: step through the same code in the Execution Visualizer, measure it in the Code Inspector, or decode an error message in Explain This Error. Everything runs in your browser; nothing you paste is uploaded or stored.

It is for beginners reading their first programs, students checking homework, developers handed code in an unfamiliar language, and anyone who has pasted an AI-generated snippet and wants to know what it actually does.

Related tools: Code Inspector, Code Execution Visualizer, and Explain This Error.