diff options
Diffstat (limited to '')
| -rw-r--r-- | dhall_parser/src/dhall.abnf | 90 |
1 files changed, 44 insertions, 46 deletions
diff --git a/dhall_parser/src/dhall.abnf b/dhall_parser/src/dhall.abnf index 2fad4d4..ca54215 100644 --- a/dhall_parser/src/dhall.abnf +++ b/dhall_parser/src/dhall.abnf @@ -132,6 +132,7 @@ whitespace-chunk = whsp = *whitespace-chunk
+; nonempty whitespace
whsp1 = 1*whitespace-chunk
; Uppercase or lowercase ASCII letter
@@ -142,19 +143,13 @@ DIGIT = %x30-39 ; 0-9 HEXDIG = DIGIT / "A" / "B" / "C" / "D" / "E" / "F"
-; A simple label cannot be one of the following reserved keywords:
-;
-; * if
-; * then
-; * else
-; * let
-; * in
-; * as
-; * using
-; * merge
-; * missing
-; * Infinity
-; * Some
+; A simple label cannot be one of the reserved keywords
+; listed in the `keyword` rule.
+; A PEG parser could use negative lookahead to
+; enforce this, e.g. as follows:
+; simple-label =
+; keyword 1*simple-label-next-char
+; / !keyword (simple-label-first-char *simple-label-next-char)
simple-label-first-char = ALPHA / "_"
simple-label-next-char = ALPHA / DIGIT / "-" / "/" / "_"
simple-label = simple-label-first-char *simple-label-next-char
@@ -174,6 +169,7 @@ label = ("`" quoted-label "`" / simple-label) ; Their list can be found in semantics.md. This is not enforced by the grammar but
; should be checked by implementations. The only place where this restriction applies
; is bound variables.
+; A PEG parser could use negative lookahead to avoid parsing those identifiers.
nonreserved-label = label
; An any-label is allowed to be one of the reserved identifiers.
@@ -216,8 +212,8 @@ any-label = label ; > "\uD834\uDD1E".
double-quote-chunk =
interpolation
- ; '\'
- / %x5C double-quote-escaped
+ ; '\' Beginning of escape sequence
+ / %x5C double-quote-escaped
/ double-quote-char
double-quote-escaped =
@@ -273,7 +269,6 @@ single-quote-char = single-quote-literal = "''" end-of-line single-quote-continue
-; Interpolation
interpolation = "${" complete-expression "}"
text-literal = (double-quote-literal / single-quote-literal)
@@ -281,25 +276,25 @@ text-literal = (double-quote-literal / single-quote-literal) ; RFC 5234 interprets string literals as case-insensitive and recommends using
; hex instead for case-sensitive strings
;
-; If you don't feel like reading hex, these are all the same as the rule name,
-; except without the '' ending.
+; If you don't feel like reading hex, these are all the same as the rule name.
; Keywords that should never be parsed as identifiers
-if = %x69.66
-then = %x74.68.65.6e
-else = %x65.6c.73.65
-let = %x6c.65.74
-in = %x69.6e
-as = %x61.73
-using = %x75.73.69.6e.67
-merge = %x6d.65.72.67.65
-missing = %x6d.69.73.73.69.6e.67
-Infinity = %x49.6e.66.69.6e.69.74.79
+if = %x69.66
+then = %x74.68.65.6e
+else = %x65.6c.73.65
+let = %x6c.65.74
+in = %x69.6e
+as = %x61.73
+using = %x75.73.69.6e.67
+merge = %x6d.65.72.67.65
+missing = %x6d.69.73.73.69.6e.67
+Infinity = %x49.6e.66.69.6e.69.74.79
+NaN = %x4e.61.4e
+Some = %x53.6f.6d.65
+
; Reserved identifiers, only needed for some special cases of parsing
Optional = %x4f.70.74.69.6f.6e.61.6c
Text = %x54.65.78.74
List = %x4c.69.73.74
-NaN = %x4e.61.4e
-Some = %x53.6f.6d.65
combine = %x2227 / "/\"
combine-types = %x2A53 / "//\\"
@@ -363,6 +358,8 @@ quoted-path-component = 1*quoted-path-character path-component = "/" ( unquoted-path-component / %x22 quoted-path-component %x22 )
+; The last path-component matched by this rule is referred to as "file" in the semantics,
+; and the other path-components as "directory".
path = 1*path-component
local =
@@ -513,7 +510,7 @@ posix-environment-variable-character = import-type = missing / local / http / env
-hash = %x73.68.61.32.35.36.3a 64HEXDIG ; "sha256:XXX...XXX"
+hash = %x73.68.61.32.35.36.3a 64HEXDIG ; "sha256:XXX...XXX"
import-hashed = import-type [ whsp hash ]
@@ -522,12 +519,6 @@ import-hashed = import-type [ whsp hash ] ; "env:FOO"
import = import-hashed [ whsp as whsp1 Text ]
-; NOTE: Every rule past this point should only reference rules that end with
-; whitespace. This ensures consistent handling of whitespace in the absence of
-; a separate lexing step.
-; The exception is the rules ending in , which should _not_ end in whitespace.
-; This is important to avoid the need for sequential backtracking in application-expression.
-
expression =
; "\(x : a) -> b"
lambda whsp "(" whsp nonreserved-label whsp ":" whsp1 expression whsp ")" whsp arrow whsp expression
@@ -544,6 +535,7 @@ expression = / forall whsp "(" whsp nonreserved-label whsp ":" whsp1 expression whsp ")" whsp arrow whsp expression
; "a -> b"
+ ;
; NOTE: Backtrack if parsing this alternative fails
/ operator-expression whsp arrow whsp expression
@@ -554,6 +546,7 @@ expression = ; "[] : List t"
; "[] : Optional t"
; "[x] : Optional t"
+ ;
; NOTE: Backtrack if parsing this alternative fails since we can't tell
; from the opening bracket whether or not this will be an empty list or
; a non-empty list
@@ -562,11 +555,17 @@ expression = ; "x : t"
/ annotated-expression
-; "x : t"
+; Nonempty-whitespace to disambiguate `env:VARIABLE` from type annotations
annotated-expression = operator-expression [ whsp ":" whsp1 expression ]
+; "let x = e1"
let-binding = let whsp1 nonreserved-label whsp [ ":" whsp1 expression whsp ] "=" whsp expression whsp
+
+; "] : List t"
+; "] : Optional t"
empty-collection = "]" whsp ":" whsp1 (List / Optional) whsp import-expression
+
+; "x] : Optional t"
non-empty-optional = expression whsp "]" whsp ":" whsp1 Optional whsp import-expression
operator-expression = import-alt-expression
@@ -592,14 +591,13 @@ not-equal-expression = application-expression *(whsp "!=" whsp application application-expression =
import-expression *(whsp1 import-expression)
-import-expression =
- import
- / selector-expression
+import-expression = import / selector-expression
; `record.field` extracts one field of a record
+;
; `record.{ field0, field1, field2 }` projects out several fields of a record
;
-; NOTE: Backtrack when parsing the `*(dot ...)`. The reason why is that you
+; NOTE: Backtrack when parsing the `*("." ...)`. The reason why is that you
; can't tell from parsing just the period whether "foo." will become "foo.bar"
; (i.e. accessing field `bar` of the record `foo`) or `foo./bar` (i.e. applying
; the function `foo` to the relative path `./bar`)
@@ -646,14 +644,13 @@ literal-expression = minus-infinity-literal = "-" Infinity
plus-infinity-literal = Infinity
-; "{ foo = 1 , bar = True }"
-; "{ foo : Integer, bar : Bool }"
record-type-or-literal =
empty-record-literal
/ non-empty-record-type-or-literal
/ empty-record-type
empty-record-literal = "=" whsp
empty-record-type = ""
+
non-empty-record-type-or-literal =
any-label whsp (non-empty-record-literal / non-empty-record-type)
non-empty-record-type = ":" whsp1 expression whsp *("," whsp record-type-entry)
@@ -666,7 +663,9 @@ record-literal-entry = any-label whsp "=" whsp expression whsp union-type-or-literal =
non-empty-union-type-or-literal
/ empty-union-type
+
empty-union-type = ""
+
non-empty-union-type-or-literal =
any-label whsp
( "=" whsp expression whsp union-type-entries
@@ -683,6 +682,5 @@ non-empty-list-literal = "[" whsp expression whsp *("," whsp expression whsp) "] parenthesized-expression = "(" whsp expression whsp ")"
-; All expressions end with trailing whitespace. This just adds a final
-; whitespace prefix for the top-level of the program
+; This just adds surrounding whitespace for the top-level of the program
complete-expression = whsp expression whsp
|