2.2. Syntax

This chapter describes how program texts are syntactically analyzed.

2.2.1. Steps of analysis

A program text is a unicode string.

A program text is first sliced to a sequence of lexemes using longest possible matching from the start to the end.

Next the lexeme sequence is reduced to a sequence of tokens, filtering out non-token lexemes.

And then, the token sequence is parsed as a program abstract source tree.

2.2.2. Lexemes

Whitespace and comments

Regular expression Lexeme category
[\u0020\u000d\u000a] Whitespace
#[^\u000a]* Comment

Whitespace characters are used to make a program text easy to read and to separate tokens from each other. Space (U+0020), carriage return (U+000d) and line feed (U+000a) are whitespace characters.


Horizontal tab (U+0009) is not a valid whitespace character. This design decision is intended to avoid bikeshed discussion about indentation.

The categories of lexemes (, [ and { may differ when they are located after whitespace characters. See the description of terminal symbols for detail.

Comments are used to describe the program text. A number sign # (U+0023) indicates a start of a comment. The comment continues till the end of the line, which is before the line feed character or the end of the program text. Comments are treated as whitespace characters.

stdout.print_line('foo'*2)      # => foofoo

stdout.print_line( 'foo' * 2 )  # => foofoo

# Comment line
do_something  # trailing comment

Whitespace characters and comments are not tokens.

Symbol tokens

Regular expression Lexeme category
[a-z_][a-z0-9_?]* Verb symbol token
([a-z_][a-z0-9_?]*)?[A-Z][a-zA-Z0-9_?]* Noun symbol token

A symbol token consists of a leading ascii letter (a-zA-Z) or an underscore _ (U+005f), and a trailing sequence of zero or more ascii letters (a-zA-Z), ascii digits (0-9), underscores _ (U+005f) and question marks ? (U+0x3f). There are two types of symbols: verbs and nouns.

A verb symbol token is a symbol token which does not contain uppercase letters [A-Z].

Verbs are commonly used for names of variables which contain functions. The following are exmaples of verbs.

  • empty?
  • _loop
  • take_5

A noun symbol token is a symbol token which contains one or more uppercase letters [A-Z].

Nouns are commonly used for names of variables which contain ordinary values. The following are examples of nouns.

  • More_lines?
  • ArrayList_class
  • rarely_Used

Num tokens

Regular expression Lexeme category
[0-9][0-9_]*(\.[0-9][0-9_]*)?(?![a-zA-Z0-9_?]) Num token (base10)
0b_*[01][01_]*(?![a-zA-Z0-9_?]) Num token (base2)
0x_*[0-9a-f][0-9a-f_]*(?![a-zA-Z0-9_?]) Num token (base16)

There are three types of num tokens: base10, base16 and base2. A base10 num token represents a num in decimal, base16 in hexadecimal, and base2 in binary.

The fractional part can be represented only by base10 num tokens. Digits before the period . (U+002e) represents the integer portion, and digits after the period represents the fractional portion.

Underscore characters _ (U+005f) can be placed for spacing.

These are examples of num tokens, each of which represents 42.

  • 42
  • 42__
  • 0042
  • 0x2a
  • 0b_10_1010

These are examples of num tokens with the fractional portion.

  • 0.0
  • 0.001
  • 3.141_592_653

All base10, base2 and base16 num tokens cannot be directly followed by a character which can form a symbol. This limitation is represented as (?![a-zA-Z0-9_?]). Thus, for example, a code fragment 24h causes a syntax error.


Without this rule, for example, 0b123 is parsed as 0b1 and 23. It is certainly error prone.

Each num token represents a number the scale of which is the count of digits of the fractional portion, and the mantissa of which is the integer made of the digits.

String tokens

Regular expression Lexeme category
'(''|[^'])*' String token (simple)
"([^"\\]|\\[0abtnvfre"\\]|\\x\{(0[0-9a-f]{0,5}|[2-9a-f][0-9a-f]{0,4}|10[0-9a-f]{0,4}|1[0-9a-f]{0,4})\})*" String token (rich)

There are two types of string tokens: simple and rich.

In a simple string token, any characters between the two single quotation marks ' (U+0027) are the content of the string. If you want to include a quotation mark itself in the string, put two consecutive quotation marks.

These are examples of simple strings.

  • 'Hello world'
  • 'Let''s go!' (it represents "Let's go!")

In a rich string token, characters between the two double quotation marks " (U+0022) are the content of the string. In the token, a sequence of characters prefixed by a backslash \ (U+005c) represents a special character, such as a line feed (\n) or a double quotation mark (\").

These are examples of rich string tokens.

  • "Let's go!"
  • "GET /index.html HTTP/1.1\r\nHost: host.example.org\r\n"

Here is a list of backslash notations.

Notation Unicode Description
\0 U+0000 Null character
\a U+0007 Bell
\b U+0008 Backspace
\t U+0009 Horizontal tab
\n U+000a Line feed
\v U+000b Vertical tab
\f U+000c Form feed
\r U+000d Carriage return
\e U+001b Escape
\" U+0022 Double quotation mark "
\\ U+005c Backslash \
\x{xxxxxx} U+xxxxxx Character specified by Unicode. xxxxxx are one to six hexadecimal digits (0-9a-f). The digits must be less than or equal to 10ffff.

Mark tokens

Tokens other than described above are called genetically mark tokens.

Some lexemes are reduced to different mark tokens depending on the conditions.

There are tokens preceding the current lexeme, and one or more whitespace characters are placed between the previous token and the current lexeme.
not WS
There are not tokens preceding the current lexeme, or no whitespace characters is placed between the previous token and the current lexeme.
Pattern Condition Mark token Note
literal ! always ! op_lognot operator
literal ~ always ~ op_not operator
literal <- always <- op_store operator
literal || always || op_logor operator
literal && always && op_logand operator
literal == always == op_eq operator
literal != always != not-equal operator
literal < always < op_lt operator
literal > always > greater-than operator
literal <= always <= less-than-or-equal-to operator
literal >= always >= greater-than-or-equal-to operator
literal | always | op_or operator
literal ^ always ^ op_xor operator
literal & always & op_and operator
literal << always << op_shl operator
literal >> always >> op_shr operator
literal + always + op_add operator
literal - always - op_sub operator, or op_minus operator
literal * always * op_mul operator
literal / always / op_div operator
literal // always // op_intdiv operator
literal % always % op_rem operator
literal = always = Let clauses
literal : not WS COLON Local variable references
literal : WS WS_COLON Attributional variable references
literal $ not WS DOLLAR Local dereference of verb variables
literal $ WS WS_DOLLAR Attributional dereference of verb variables
literal . always . Access to variables or functions
literal !! always !! Direct call expressions
literal ... always ... Elements spreading
literal [ not WS OPENBRACKET Vectors, formal receivers, or actual receivers
literal [ WS WS_OPENBRACKET Vectors
literal ] always ] Closing
literal { not WS OPENBRACE Function expressions or function arguments
literal { WS WS_OPENBRACE Function expressions
literal } always } Closing
literal ( not WS OPENPAREN Parentheses expressions, formal arguments, or actual arguments
literal ( WS WS_OPENPAREN Parentheses expressions
literal ) always ) Closing
regex \binding(?![a-zA-Z0-9_?]) always \binding Context binding expressions

2.2.3. Abstract source tree

The list describes the parsing rule of the program abstract source tree, or AST, from the token sequence of a program text.

There are several shift/reduce conflicts in the rules, and the parser always choose to shift.

program           ::=  toplevel
toplevel          ::=  empty
                       expression toplevel
seq               ::=  empty
substantial_seq   ::=  expression
                       expression substantial_seq
                       expression '=' expression substantial_seq
expression        ::=  store_op
store_op          ::=  logor_op
                       logor_op '<-' logor_op
logor_op          ::=  logand_op
                       logand_op '||' logor_op
logand_op         ::=  relation_op
                       relation_op '&&' logand_op
relation_op       ::=  add_op
                       add_op '==' add_op
                       add_op '!=' add_op
                       add_op '<' add_op
                       add_op '>' add_op
                       add_op '<=' add_op
                       add_op '>=' add_op
add_op            ::=  multiply_op
                       add_op '+' multiply_op
                       add_op '-' multiply_op
                       add_op '|' multiply_op
                       add_op '^' multiply_op
multiply_op       ::=  unary_op
                       multiply_op '*' unary_op
                       multiply_op '/' unary_op
                       multiply_op '//' unary_op
                       multiply_op '%' unary_op
                       multiply_op '&' unary_op
                       multiply_op '<<' unary_op
                       multiply_op '>>' unary_op
unary_op          ::=  primary
                       '-' unary_op
                       '!' unary_op
                       '~' unary_op
primary           ::=  num
num               ::=  NUM
str               ::=  STRING
context_binding   ::=  '\binding'
paren             ::=  OPENPAREN seq ')'
                       WS_OPENPAREN seq ')'
vec               ::=  OPENBRACKET vec_body ']'
                       WS_OPENBRACKET vec_body ']'
fun               ::=  OPENBRACE fun_body '}'
                       WS_OPENBRACE fun_body '}'
local_deref       ::=  NOUN | DOLLAR VERB | WS_DOLLAR VERB
local_ref         ::=  COLON NOUN
                       WS_COLON NOUN
                       COLON VERB
                       WS_COLON VERB
local_call        ::=  VERB recv args
attr_deref        ::=  primary '.' NOUN
                       primary DOLLAR VERB
attr_ref          ::=  primary COLON NOUN
                       primary COLON VERB
attr_call         ::=  primary '.' VERB recv args
direct_call       ::=  primary '!!' recv args
recv              ::=  empty
                       OPENBRACKET expression ']'
args              ::=  paren_args fun_args
paren_args        ::=  empty
                       OPENPAREN vec_body ')'
fun_args          ::=  empty
                       fun_arg fun_args
fun_arg           ::=  OPENBRACE fun_body '}'
vec_body          ::=  empty
                       elements_producer vec_body
elements_producer ::=  expression
                       '...' expression
fun_body          ::=  formal_receiver formal_args seq
formal_receiver   ::=  empty
                       OPENBRACKET expression ']'
formal_args       ::=  empty
                       OPENPAREN vec_body ')'
empty             ::= 

These symbols shown in the production rules are terminal symbols.

Terminal symbol Description
VERB Verb symbol such as print
NOUN Noun symbol such as Count
NUM Num literal such as 42 or 3.14159
STRING String literal such as 'Peng!'
COLON Mark token COLON
Those enclosed by single quotation marks Corresponding mark token