4.23.1 Base32 Encoding

4.23.1.1 base32-encode

(base32-encode (bytes : :u8vector)
               padding: (padding? : :boolean := #t)
               lowercase: (lowercase? : :boolean := #f))
  => :string

Returns a new string encoding all input bytes, preserving leading zero bytes. The default is uppercase with = padding to a multiple of eight characters. padding: #f omits padding; lowercase: #t uses a-z2-7. No line breaks or prefixes are inserted. Empty input returns an empty string.

For n input bytes the output length is 8 * ceiling(n / 5) with padding, or ceiling(8 * n / 5) without it. An output length outside the fixnum range raises ContractViolation; allocation failures propagate from the runtime.

4.23.1.2 base32-decode

(base32-decode (str : :string)
               no-padding: (no-padding? : :boolean := #f))
  => :u8vector

Returns a new byte vector. Uppercase, lowercase and mixed-case letters are accepted. By default, complete padding is required for partial eight-character groups. Complete groups need no padding. no-padding: #t additionally permits correct unpadded input; it still accepts correctly padded input and does not relax any other validation. Empty input is valid in either mode.

The decoder rejects:

• Characters outside A-Z, a-z, 2-7, and terminal padding, including whitespace, non-ASCII characters, 0, 1, 8, and 9.
• Data lengths congruent to 1, 3, or 6 modulo 8.
• Interior, missing (unless permitted), excessive, or incorrect padding. Data-length residues 2, 4, 5, and 7 require respectively 6, 4, 3, and 1 padding characters when padding is present. Complete groups take none.
• Nonzero unused low bits in the last data character, even without padding.

Malformed encodings raise ContractViolation with base32-decode as the error location and diagnostic irritants: input length, plus data-length for impossible lengths, index and padding for padding errors, or index and char for invalid characters and pad bits. Indexes are zero-based; padding errors report the start of trailing padding (or the input length when required padding is missing). If several rules are violated, length and padding validation can report an error before character validation. Argument types and boolean options are checked by the declared contracts.

4.23.1.3 DNS-Safe Labels

(import :std/encoding/base32)

(base32-encode #u8(102 111 111) padding: #f lowercase: #t)
;; => "mzxw6"
(base32-decode "mzxw6" no-padding: #t)
;; => #u8(102 111 111)

Lowercase unpadded output uses only DNS-safe letters and digits. The codec does not enforce DNS label or hostname limits, add DID prefixes, or split labels. A single DNS label must be nonempty and at most 63 octets; without additional prefixes, at most 39 input bytes fit in one Base32 label.

Both operations run in linear time with a bounded accumulator of at most 12 bits and constant auxiliary space besides the result. Decoding validates without re-encoding and allocates exactly floor(5 * d / 8) bytes for d data characters. Inputs are not modified. There is no application-specific size limit beyond runtime string/vector capacity, the encoding fixnum-size check, and available memory.