When this matters
Start here when source text is rejected before evaluation, or when you are unsure whether parentheses describe a call or data. The reader owns spelling and structure; evaluation has not begun yet.
See it run
LISPEX
'(alpha 1 2/3 #u8(4 5))
Observed result
OUTPUT
(alpha 1 2/3 #u8(4 5))Read the example
The leading quote keeps the whole form as data. The reader still recognizes a symbol, integer, rational, and bytevector, but none of them is called or computed. Removing the quote would put alpha in procedure position.
How to reason about it
- Integer, rational, and real token grammars are disjoint; leading-zero and non-finite numeric spellings are rejected.
- Lists, dotted pairs, vectors, bytevectors, strings, characters, quote, quasiquote, and comments have pinned forms.
- Reader failures are E1xx diagnostics with source spans; invalid UTF-8 is rejected at the invocation boundary.
Choose quickly
| You want | Write | What the reader creates |
|---|---|---|
| a name | alpha | a symbol token that may later be looked up |
| text | "alpha" | an immutable string |
| a proper list as data | '(alpha 1) | a quoted pair chain ending in () |
| a dotted pair as data | '(alpha . 1) | one pair whose tail is 1 |
| indexed values | #(1 2) | a vector |
| bytes | #u8(1 2) | an immutable bytevector |
A common mistake
Radix and exactness prefixes such as #x and #e are outside the current profile.
Current boundaries
- The language does not silently normalize newlines, paths, or host encodings.
Keep going
Continue with evaluation order to see what happens after reading, or open Reader Grammar for every accepted token shape.