Katas

Level: reference · the practice track

One line: The lessons explain; the katas make you predict. Each one lives on the page for the topic it teaches — this is the only place they are put in an order.

Where a kata lives

A kata belongs to its topic, not to a folder of its own. The page that explains the byte is the page that asks you to write one down, under a ## Practice heading near the end, with the answer folded into a <details markdown="1"> block.

That is deliberate, and the reason is that folders are URLs. A topic — the byte, the code point, grep's locale — is stable for years; a sequence is not. The moment a kata belongs between K1 and K2, a K01_…/ folder either gets renumbered, breaking every link anyone saved, or starts lying about its own order. So the sequence lives here, in a table that costs nothing to reorder, and the numbers below are labels rather than addresses. Same rule as the sidebar: order is presentation, so it belongs in a page, never in a path.

Every answer here is printed by a program, recorded as an answer key and run by CI on Ubuntu and macOS with every other example in the library — so a solution cannot rot into one that no longer says what the page says. check_katas.py enforces that, and the rules are written down in CONTRIBUTING.md.

The katas

#	Kata	Lesson	Level
K1	One byte, four readings — and the byte decides none of them	A byte is eight bits	101
K2	Turn the odometer in your head — five hex questions, one carry	Counting in hexadecimal	101
K3	Four bits at a time, both ways — then find the byte boundary in octal	Hex is a shorthand	101
K4	Read the three columns off a file whose bytes you chose	Reading a hex dump	101
K5	Thirteen bytes at three widths, and where the ragged piece falls	Grouping is a choice	101 → 201
K6	The hash field that arrived one character short, read three ways	Hex: a number, or a picture of bytes	201
K7	Two columns for four expressions — and where the loss happens in each language	Arithmetic has its own width	101 → 201
K8	Four bytes: 01 00 00 00, and which two of the four answers are the file	The bytes do not say which end	101 → 201
K9	Say which base the field is in — five questions, two brackets and one dump	Which base did you mean?	201
K10	Three ASCII tricks, and the four characters that break them	A character is a number	101
K11	Rotate, rotate back, then step outside ASCII and lose both guarantees	Rotation is not encryption	101 → 201
K12	How many lines is this string? Three answers, all correct	Control characters	101
K13	Five doors, and the one byte that is valid UTF-8 and welcome nowhere	The NUL byte	201
K14	One byte, four tables — and the byte that settles the superset question	Code pages	101 → 201
K15	Four code points read as addresses, and what each costs to write	Unicode code points	101 → 201
K16	Four spellings of é, then the character \u cannot reach	Writing a code point	201
K17	Five claims: which go in a bug report without a version stamp?	The table has a version	201
K18	Six strings, three of them one user — and when folding is the bug	Preparing a string	301
K19	Five numbers for one Polish word, and the letter that does not come apart	A code point is not a character	201
K20	Five checks that all say no, and only one for the right reason	Confusables and scripts	301
K21	Five ways to type one variable name, and the language that says so	Unicode in identifiers	301
K22	Encode four characters with a pencil, then the one that is not a character	UTF-8 by hand	101 → 201
K23	One bad byte, and how much of the check survives the call	Validation is a boundary	201 → 301
K24	Two spellings of a slash, and the filter that only sees one	Overlong sequences	301
K25	Build the surrogate pair by hand, then explain U+10FFFF	UTF-16 and surrogates	201
K26	Write the bytes for six encodings, then name the CSV column	Byte order and the BOM	201
K27	Five lines, two loud failures and two quiet wrong answers	Encode and decode are verbs	101
K28	Name the culprit from the garbage, and say which garbage is repairable	Mojibake	101 → 201
K29	One character through four channels, and what each one wraps	Escaping into ASCII	201
K30	Predict four output lengths, then find the 4/3 that is not there	Binary to text	201
K31	Four Base32 menu entries, and the one that is a different encoding	The alphabet is not the encoding	201
K32	Seven numbers for one two-character string	An encoding is four layers	201
K33	Six error policies, and the only one you can reverse	Encode, decode and errors	101 → 201
K34	Nine ways out of one string, and the two bytes it will not carry	Bytes that are not text	201
K35	Two spellings, four forms, and the lookup that raises nothing	Normalization	201
K36	Five refusals, one invariant	String is bytes that promise UTF-8	101 → 201
K37	Four numbers, one character — and the fifth, which is the first again	char is four bytes	101 → 201
K38	Three contracts for four bytes, and the byte that changes the verdict	From UTF-8, and lossy	201
K39	Where may you cut "żółw"? — five indices, and a field of five bytes	Slicing by byte	201
K40	Three doors out of one filename, and the only one that hands it back	OsStr, Path, and WTF-8	301
K41	A file with a letter in it and zero lines, and the loop that drops it	The trailing newline	101 → 201
K42	Three decisions about one newline byte, and the row that needs all three	A character and its bytes on one line	101
K43	Five readings of six bytes -- which column is the file?	Inspecting a file	101 → 201
K44	Six variables, and whether any of them changes a Python len()	Locale and LC_CTYPE	201
K45	One file, four answers, and which one the kernel consults	File type is four questions	201
K46	Three scripts, six answers, and the two that surprise you	The first two bytes	101 → 201
K47	Assemble the escape sequence, then find the byte nobody escapes	Terminal hyperlinks, and the URI that is not one	201
K48	Three bytes that are the fix and the bug at once	A BOM in a CSV	101 → 201
K49	Totals right to the cent, keys that match nothing	CRLF vs LF	101 → 201
K50	Five scars, and the constraint that made each one right	From the telegraph to Unicode	201
K51	Six properties in six lines, and the bill it pays	Why UTF-8 won	201
K52	What those four systems actually adopted, and why they cannot leave	Why UTF-16 stayed	201
K53	Five rules, and the line of code each one forbids	UTF-8 everywhere	201
K54	Five signatures, and the one the compiler refuses	Rust strings in practice	201 → 301
K55	Four habits, and the bytes method that is right and still a bug	Python text in practice	201
K56	Six round trips, and the three that come back unchanged	"Handles Unicode" is four questions	301
K57	What does a dot match? Four answers, and who decided	grep on text that is not ASCII	201
K58	Three marks, one missing encoding, and three defaults that hide a file	ripgrep — the Rust grep	201
K59	The only engine that can match one grapheme, and what it stops telling you	PCRE2 — the other regex engine	201
K60	Two stages, and why removing a container fixes no encoding	--pre and -z — decompress, then decode	201
K61	A file cat can open and find -name cannot see	find, and filenames that are bytes	201
K62	Four filenames, and how many survive the naive pipeline	xargs splits on the wrong things	201
K63	The edit that works, the locale that breaks it, and the -i with no portable spelling	sed matches patterns, not bytes	201
K64	Whose awk is this, and what does length() count?	awk is three programs	201
K65	How long is this variable, and which two bytes never arrive?	The shell has no string type	201
K66	-b or -c, and the machine that changes the answer	cut counts what it is told to count	201
K67	Delete one letter, damage two words -- sets against sequences	tr and sort work a byte at a time	201
K68	Four ways to ask if two files match, none of them the question you meant	diff compares lines, cmp compares bytes, neither compares text	201
K69	Three tools that take text as bytes, and the one with no opinion	split cuts characters, paste cuts delimiters, look needs a sorted file, tee does nothing	201
K70	Seven ways to blank a file, and the one that leaves the old bytes behind	touch, : > and install are not three spellings of one command	201
K71	Why the pairs came out swapped, and what a grouping width claims	hexdump is a format engine wearing six presets	201
K72	Edit the dump and put it back -- and find the column that was ignored	xxd is the dump you can put back	201
K73	Read the interface as a C type, and say why -a cannot be quoted	od reads types, not bytes	201
K74	Six columns, and the one no dump tool has	uni — the character's name	101 → 201
K75	Predict the row, then name the tables, then read a bare number	uni -h, line by line	201
K76	Four ways to type it, and the two that identify the character	Typing a character you cannot type	101 → 201
K77	One byte string, two readers, and a duplicate JSON key	Two readers, one byte string	301
K78	Five paths, two orders, and why checking harder is not the fix	The check that ran too early	301
K79	Six signups, three classes, and who owns each one	Two people, one account	301
K80	Five bytes that end a field for somebody else	The byte that means something to somebody else	301
K81	Two readers of one file, and the one with no access to the bytes	What you see is not what runs	301
K82	Forty pages nobody opens, and how to find them without the names	The encoding man pages nobody opens	201
K83	Out of date, or wrong? -- and the three dates a page carries	A page has a date	201
K84	The flag no page mentions, and the two platforms that disagree about it	What the page does not say	201
K85	Five damaged fields, and the question that sorts them	The mojibake round trip	201
K86	Three ways to make one string wider, and the four calls that move a number	A str in memory	201
K87	One file, four readings, and the same script on three machines	Opening a file	201
K88	Six bytes, and a format spec you have not been given	Bytes, hex and int	201
K89	Four iconv commands over one six-byte file	iconv	201
K90	Four writes that look identical, and the one character that breaks the tie	printf writes bytes	101
K91	Three files, one verdict, and only one of them is ASCII	file guesses	101
K92	The same script, two commands, and the line that never arrives	A pipe is not a terminal	201
K93	Five code points, and the one a UTF-8 encoder actually refuses	Noncharacters and the private use areas	201
K94	Five byte strings, one tag, and the one an encoder will never write	UTF-7, and the seven-bit transport	201 → 301
K95	Four records, one lie — and the number the checksum hands back	A record has to say what it is, how long it is, and whether it arrived	201 → 301
K96	Six questions about fourteen bytes, and the one with no answer	Packing a record	201 → 301
K97	Six expressions, and the requirement each one is about	"Supports Unicode" is a level, not a yes	301
K98	Four pairs, and the one no level can separate	Sorting and collation	301
K99	Five pairs, and the checks that reach them	The hard strings	301
K100	Four files, twelve answers, and a rename	Binary is a verdict, not a property	101
K101	Five fields cut to fit, and the one a Go front end refuses	rune is an int32	201