Repository navigation
Front end: := on a UDT field in a single-line function body, and CRLF line endings #230
Description
Summary
Two small front-end gaps that stop otherwise valid scripts before analysis.
Versions
pineforge-codegen 1.6.1
1. Field reassignment in a single-line function body
//@version=6
strategy('Field assignment')
type State
int value = 0
set(State s) => s.value := 1
var State s = State.new()
set(s)
plot(s.value)
PARSER 5:25 Unexpected token COLON_EQUALS(':=') after statement; expected a line break or comma
The same body written on its own indented line translates.
2. CRLF line endings
The three-line script //@version=6, strategy('Newlines'), plot(close) translates with LF endings and fails with CRLF:
LEXER 2:21 Unexpected character: '\r' (and on every following line)
Scripts saved by Windows editors arrive with CRLF. Accepting \r\n (and a lone \r) as a line break would avoid every host having to normalise.
Activity
letiep97 commentedon Oct 20, 2026
Both still reproduce on main (
aebad8c). Pointers in case they help:_parse_func_defand_parse_method_defread a single expression after=>(parser.py#L921-L932, #L1050-L1061)._parse_arm_line(parser.py#L1204) already parses a one-line block, including a field target such aso.f := v, for switch arms. Using it for one-line function and method bodies should cover this; whatever readsis_single_exprmay need a look.normalize_library_text, library_modules.py#L61-L67). Applying the same normalization to the main source at the top of_parse_bounded(init.py#L25-L33), before the pragma scan and the lexer, would cover it. Diagnostic line numbers don't change for\r\n.luisleo526 commentedon Oct 25, 2026
Thank you for the report. Part 2 (CRLF): confirmed in the v1.7.0 source,
\ris rejected by the lexer. This is not fixed in v1.7.0. A fix is being written:\r\nand a lone\rwill be treated as a line break, with the same output and the same line numbers as the LF form. We expect to open the pull request on 2026-10-26. Part 1 (:=on a UDT field in a single-line function body): confirmed in the v1.7.0 source and not fixed in v1.7.0; we plan to open a pull request for it on 2026-10-30.luisleo526 commentedon Oct 25, 2026
Update on part 2 (CRLF): the fix is up for review in #252. One precision to my earlier note:
\r\nis folded to\neverywhere, including inside strings, and a lone\rends a line in code and comments; a lone\rinside a string literal is kept as part of the string. Part 1 (:=on a UDT field in a single-line function body) is still planned for a pull request on 2026-10-30.test: give the double-CRLF scripts a version directive (#230)
codegen: CRLF and lone CR line endings (#230, line-ending part) (#252)
luisleo526 commentedon Oct 25, 2026
Part 2 (CRLF / lone CR line endings) is fixed by #252, merged to main; it ships in the next codegen release, 1.8.0 (I will confirm here when it is published).
Reopening, because part 1 (
:=on a UDT field in a single-line function body) is not fixed yet: the merge closed this issue automatically. Part 1 is still planned for a pull request on 2026-10-30.docs: CHANGELOG: the planned 1.8.0 section with migration notes (#257)
codegen: five fixes for 1.8.0 (late restore, statement-order var init…
luisleo526 commentedon Oct 26, 2026
Both parts of this issue are fixed on main:
\r\nand a lone\rare line breaks, with the same output and line numbers as the LF form.:=on a UDT field in a single-line function or method body): codegen: five fixes for 1.8.0 (late restore, statement-order var initialization, request.currency_rate, helper-local var color, one-line field assignment) #261 (the parser change originally proposed as parser: a UDT field assignment as a one-line function or method body (Part of #230) #253).set(State s) => s.value := 1now translates like the same body on its own indented line. A plain-name or subscript target in a one-line body keeps its parser error, as before.Both ship in codegen 1.8.0. Closing this issue as fixed on main; I will confirm here when 1.8.0 is published.
luisleo526 commentedon Oct 26, 2026
Both parts of this issue are fixed on main:
\r\nand a lone\rare line breaks, with the same output and line numbers as the LF form.:=on a UDT field in a single-line function or method body): codegen: five fixes for 1.8.0 (late restore, statement-order var initialization, request.currency_rate, helper-local var color, one-line field assignment) #261 (the parser change originally proposed as parser: a UDT field assignment as a one-line function or method body (Part of #230) #253).set(State s) => s.value := 1now translates like the same body on its own indented line. A plain-name or subscript target in a one-line body keeps its parser error, as before.Both ship in codegen 1.8.0. Closing this issue as fixed on main; I will confirm here when 1.8.0 is published.
letiep97 commentedon Oct 28, 2026
Verified both parts on main (
11b5ca5, with engine 1.6.1): the:=on a UDT field in a one-line function body now translates and runs, and the CRLF file from this issue now translates. Thank you for the fixes.