Source text and lexical elements

Encoding#

A source file is UTF-8 text of at most 64 MiB. Lines end with LF or CRLF; a leading byte-order mark is ignored.

What is not text is refused before parsing begins, once, naming the file and the position inside it.

Bidirectional control characters are refused everywhere.

The program may come from standard input (luce check -, or any pipe). Diagnostics then name it <stdin>, and imports resolve beside the current directory.

Indentation#

Blocks are introduced by : at the end of a line and delimited by indentation. One level is four spaces, exactly. Tabs are not indentation and neither is any other width.

Comments#

# to the end of the line. There is no block comment form.

Identifiers#

[A-Za-z_][A-Za-z0-9_]*. Case is significant. Convention, which the compiler relies on for type names, is TitleCase for types and snake_case for everything else.

There is no shadowing: a name declared in an enclosing scope cannot be re-declared in an inner one.

Keywords#

and      break    catch    continue copy     elif     else
false    for      func     give     if       import   in
let      new      none     not      or       return   struct
true     try      var      while

Reserved names#

The language reserves these; nothing user-declared may take them.

input   output   Input   output   range   Int     Float   Bool
String  Bytes    List    Map      Array   Builder None    abs
min     max      clamp   sqrt     floor   ceil    len     slice
byte_at assert   trap    str      parse_int        parse_float
chr     ord      append  pop      insert  remove  has     dim
free    print    file_read        file_write       file_exists
arg     arg_count         key_read         key_text         error

Number literals#

Decimal only. An integer literal is a sequence of digits and yields an Int; a fraction or an exponent yields a Float.

12        Int
1.5       Float
1e10      Float
1.5e-3    Float

A . starts a fraction only when a digit follows it. There are no hexadecimal, binary or octal literals and no _ digit separators — writing one is luce.lex.number, naming the reason, rather than a silent misreading. A non-finite float literal such as 1e400 is refused.

String literals#

"...", on one line. The escapes are exactly four:

Every other escape is rejected by name — \r, \0, and hex and unicode escapes included. A codepoint goes in with chr(...).

An f"..." literal is an interpolated string: {expression} holes are converted with str(...) and concatenated. {{ and }} are literal braces. A hole holds one expression, and "..." strings nested inside a hole are permitted.

Operators and punctuation#

+  -  *  /  %          arithmetic
== != <  <= >  >=      comparison (non-associative)
=  += -= *= /= %=      assignment
:  ,  .  ->  ?  !      declaration and type syntax
(  )  [  ]  _          grouping, indexing, array shape

Diagnostics#

Every diagnostic carries a stable code and a byte span, and renders as file:line:column in both build modes. Codes are namespaced by the stage that raised them: luce.source.*, luce.lex.*, luce.parse.*, luce.sema.*, luce.import.*.

At most 100 diagnostics are reported for one compilation.