1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
|
Parsing
=======
Elymas has a very simplistic parser (if at all). Parsing works as follows:
* Spaces separate things.
* If a `#` is encountered, ignore rest of line. This implements comments.
* If a sequence of digits is encountered, this is an integer literal and gets
pushed to the stack.
* If something looks like a float, i.e. matches (\\d[0-9.]*([eE]-?\\d+)?),
this is a float literal and gets pushed to the stack.
* If a `"` is encountered, a string literal begins. Within a string, backslash
escapes special characters. `\\` becomes single backslash, `\n` is a new line,
`\r` is a carriage return, `\0` is the NUL character (all bits zero). All
other (non-escaped) characters just become part of the string.
* If a sequence of alphanumeric characters (a-zA-Z0-9) is encountered,
it is looked up in the current scope. This is how normal functios like `dump`
get referenced.
* If a sequence of non-alphanumeric characters is encountered,
it is looked up in the current scope. This is how functions like `_` work.
* If a sequence starts with non-alphanumeric characters, but then an alphanumeric
character is encountered, this latter part is converted to a string literal
(backslash escapes do not apply) and pushed to the stack. Afterwards the
non-alphanumeric prefix is looked up in the current scope. This is how
`/abc` becomes `"abc"`: The string is created during parsing and afterwards
the identity function is applied.
Additionally, the parser has a quote-level counter which is increased by `{`
and decreased by `}`. If the quote-level is non-zero, i.e. the parser is in
quote mode, the resolution of names is postponed and a function object doing
the resolution is pushed instead.
|
