[
https://issues.apache.org/jira/browse/GROOVY-9894?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17264519#comment-17264519
]
David Stafford commented on GROOVY-9894:
----------------------------------------
Looking at it harder, I wonder how it ever worked. I now see the /$ then a $/
then a $ followed by the rest of the document, which doesn't end until we
happen to bump into another /$ much later. How the old parser parsed it the
way I thought (and Eclipse syntax highlighting also marks it as ending after
+)/$$/$.
So my mistake, although there's an issue in the 2.5 parser. My fix is
static final String uriRegex = $/^/([a-zA-Z0-9]+)$/$$/$
or possibly
static final String uriRegex = $/^/([a-zA-Z0-9]+)$/$/$
Captain Hindsight says it's too bad the syntax wasn't @/^/([a-zA-Z0-9]+)/$/@
since @ isn't used in any regex I know of.
> Dollar slashy string ending in escaped dollar parses in 2.0 fails in 3.0
> ------------------------------------------------------------------------
>
> Key: GROOVY-9894
> URL: https://issues.apache.org/jira/browse/GROOVY-9894
> Project: Groovy
> Issue Type: Bug
> Components: lexer
> Affects Versions: 3.0.5, 3.0.7
> Environment: Windows 10
> Reporter: David Stafford
> Priority: Major
>
> This program fails to compile in 3.0.7 (and 3.0.5) but parses in 2.5:
> {code:java}
> public class TestClass {
> static final String uriRegex = $/^/([a-zA-Z0-9]+)/$$/$
> public TestClass(
> ) {
> // String uuid = UUID.randomUUID().toString();
> // "/${evenme}/${uuid}.xml"
> }
> }
> {code}
> error is:
> {noformat}
> org.codehaus.groovy.control.MultipleCompilationErrorsException: startup
> failed:
> src\main\groovy\com\broadridge\datafabric\publisher\enrichment\formatters\TestClass.groovy:
> 7: Unexpected input: '$/$\r\n\r\n\tpublic TestClass(\r\n\t)
> {\r\n\t\t//\t\tString uuid = UUID.randomUUID().toString();\r\n //
> "/${evenme}/${uuid}.xml"' @ line 7, column 35.
> // "/${evenme}/${uuid}.xml"
> ^
> 1 error
> {noformat}
> In my actual code, only the comment is pointed out as incorrect, which was
> very disconcerting. The issue is the $/$ in uriRegex is taking precedence
> over the $$ (possibly because its a longer but later match?). This means the
> lexer thinks we're still in a dollar-slashy string. What it should do is
> consume the $$, turn it into a $ (escaped dollar), then hit the /$, and end
> the string.
> My workaround is to convert to a slashy string, but that means I need to
> escape all the forward slashes in the regex, and that's what dollar slashy
> strings are good for (avoiding extra escapes). As it stands, in 3.0 a
> dollar-slashy can't end with a dollar, which is a pretty common character for
> a regex to end with.
> This came up when converting a large codebase to groovy 3.0, getting bizarre
> error messages pointing far away from the actual issue.
>
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
