Number Base Converter

Every value is arbitrary precision, and that is the point

The standard parsing functions lose precision above two to the fifty-third, and the values people bring to a base converter are routinely above it: a 64-bit snowflake id, a nanosecond timestamp, a permissions bitmask, a hash prefix. A tool that silently rounds a sixteen-digit integer gives you a confidently wrong answer, which is worse than refusing to answer.

Everything here uses arbitrary-precision integers, so a 64-bit value converts exactly in every base. Underscores, spaces and commas in the input are ignored, so a grouped literal copied out of source code pastes straight in.

The signed reading is the reason to use this

A base table is one library call. What is worth showing is that the same byte is 255 unsigned and minus one signed, and that which of those your program sees depends on a type declaration nowhere near the literal.

So the bit pattern is rendered as a two’s-complement signed integer at every width it fits in, and the widths where that reading differs from the unsigned value are marked - because those are exactly the cases where the top bit is set and somebody is about to be surprised.

• Binary grouped into nibbles, so bit positions can be read off the page
• Which bit positions are set, for a flag set or a permissions mask
• The value as big-endian bytes, matching what a hex dump shows
• A warning above the safe-integer limit, where the platform’s own number type stops being exact

Base detection

With detection on, a hex, binary or octal prefix selects the base and anything else is read as decimal - which is what a developer pasting a value out of a log or a source file expects, and it means the common case needs no configuration at all. An explicit base still tolerates a matching prefix, so pasting a prefixed hex value with hex already selected is not an error.

What it will not do

It handles integers only. Floating-point values have their own representation - sign, exponent and mantissa - and showing a float’s bit pattern is a different tool with a different explanation, so a decimal point is rejected rather than truncated.

It does not perform bitwise arithmetic between two values, and it does not know what your bits mean. A permissions mask, a feature-flag set and a colour value are all the same number here; naming the bits is your context, not the converter’s.