feat: detect infinite recursion

Author: Aloso

pomsky-lib/src/diagnose/compile_error.rs

@@ -95,6 +95,7 @@ pub(crate) enum CompileErrorKind {
         flavor: RegexFlavor,
     },
     NestedTest,
+    InfiniteRecursion,
     BadIntersection,
     EmptyIntersection,
 }
@@ -219,6 +220,7 @@ impl core::fmt::Display for CompileErrorKind {
                 ),
                 _ => write!(f, "This kind of lookbehind is not supported in the {flavor:?} flavor"),
             },
+            CompileErrorKind::InfiniteRecursion => write!(f, "This recursion never terminates"),
             CompileErrorKind::BadIntersection => write!(
                 f,
                 "Intersecting these expressions is not supported. Only character sets \

pomsky-lib/src/diagnose/diagnostic_code.rs

@@ -98,8 +98,9 @@ diagnostic_code! {
         RubyLookaheadInLookbehind = 319,
         UnsupportedInLookbehind = 320,
         LookbehindNotConstantLength = 321,
-        BadIntersection = 322,
-        EmptyIntersection = 323,
+        InfiniteRecursion = 322,
+        BadIntersection = 323,
+        EmptyIntersection = 324,
 
         // Warning indicating something might not be supported
         PossiblyUnsupported = 400,
@@ -234,6 +235,7 @@ impl<'a> From<&'a CompileErrorKind> for DiagnosticCode {
             C::UnsupportedInLookbehind { .. } => Self::UnsupportedInLookbehind,
             C::LookbehindNotConstantLength { .. } => Self::LookbehindNotConstantLength,
             C::NestedTest => Self::NestedTest,
+            C::InfiniteRecursion => Self::InfiniteRecursion,
             C::BadIntersection => Self::BadIntersection,
             C::EmptyIntersection => Self::EmptyIntersection,
         }

pomsky-lib/src/diagnose/diagnostic_kind.rs

@@ -40,9 +40,10 @@ impl From<&CompileErrorKind> for DiagnosticKind {
             | K::NameUsedMultipleTimes(_)
             | K::UnknownVariable { .. }
             | K::RelativeRefZero => DiagnosticKind::Resolve,
-            K::EmptyClassNegated { .. } | K::IllegalNegation { .. } | K::EmptyIntersection => {
-                DiagnosticKind::Invalid
-            }
+            K::EmptyClassNegated { .. }
+            | K::InfiniteRecursion
+            | K::IllegalNegation { .. }
+            | K::EmptyIntersection => DiagnosticKind::Invalid,
             K::CaptureInLet
             | K::ReferenceInLet
             | K::RecursiveVariable

pomsky-lib/src/diagnose/help.rs

@@ -180,6 +180,11 @@ pub(super) fn get_compiler_help(kind: &CompileErrorKind, _span: Span) -> Option<
             Parentheses may be required to clarify the parsing order."
                 .to_string(),
         ),
+        CompileErrorKind::InfiniteRecursion => Some(
+            "A recursive expression must have a branch that \
+            doesn't reach the `recursion`, or can repeat 0 times"
+                .to_string(),
+        ),
 
         _ => None,
     }

pomsky-lib/src/exprs/mod.rs

@@ -1,7 +1,7 @@
 use crate::{
     capturing_groups::CapturingGroupsCollector,
     compile::{CompileResult, CompileState},
-    diagnose::Diagnostic,
+    diagnose::{CompileErrorKind, Diagnostic},
     options::CompileOptions,
     regex::Count,
     validation::Validator,
@@ -60,7 +60,8 @@ impl Expr {
         input: &str,
         options: CompileOptions,
     ) -> (Option<String>, Vec<Diagnostic>) {
-        if let Err(e) = Validator::new(options).visit_rule(&self.0) {
+        let mut validator = Validator::new(options);
+        if let Err(e) = validator.visit_rule(&self.0) {
             return (None, vec![e.diagnostic(input)]);
         }
 
@@ -90,6 +91,12 @@ impl Expr {
             Ok(compiled) => compiled,
             Err(e) => return (None, vec![e.diagnostic(input)]),
         };
+        if let Some(rec_span) = validator.first_recursion {
+            if !compiled.terminates() {
+                let error = CompileErrorKind::InfiniteRecursion.at(rec_span);
+                return (None, vec![error.diagnostic(input)]);
+            }
+        }
         let count = compiled.optimize();
 
         let mut buf = String::new();

pomsky-lib/src/regex/mod.rs

@@ -156,6 +156,19 @@ impl Regex {
             matches!(self, Regex::CharSet(_))
         }
     }
+
+    pub(super) fn terminates(&self) -> bool {
+        match self {
+            Regex::Recursion => false,
+            Regex::Repetition(repetition) => {
+                repetition.kind.lower_bound == 0 || repetition.content.terminates()
+            }
+            Regex::Group(group) => group.parts.iter().all(|part| part.terminates()),
+            Regex::Alternation(alternation) => alternation.parts.iter().any(|alt| alt.terminates()),
+            Regex::Lookaround(lookaround) => lookaround.content.terminates(),
+            _ => true,
+        }
+    }
 }
 
 impl Default for Regex {

pomsky-lib/src/validation.rs

@@ -10,12 +10,13 @@ use crate::{
 #[derive(Clone)]
 pub(crate) struct Validator {
     pub(crate) options: CompileOptions,
+    pub(crate) first_recursion: Option<Span>,
     pub(crate) layer: u32,
 }
 
 impl Validator {
     pub(crate) fn new(options: CompileOptions) -> Self {
-        Validator { options, layer: 0 }
+        Validator { options, first_recursion: None, layer: 0 }
     }
 
     fn require(&self, feature: u16, span: Span) -> Result<(), CompileError> {
@@ -132,6 +133,10 @@ impl RuleVisitor<CompileError> for Validator {
     fn visit_recursion(&mut self, recursion: &exprs::Recursion) -> Result<(), CompileError> {
         self.require(Feat::RECURSION, recursion.span)?;
 
+        if self.first_recursion.is_none() {
+            self.first_recursion = Some(recursion.span);
+        }
+
         if let RegexFlavor::Pcre | RegexFlavor::Ruby = self.flavor() {
             Ok(())
         } else {

pomsky-lib/tests/testcases/recursion/never_terminates.txt

@@ -0,0 +1,6 @@
+#! expect=error, flavor=Pcre
+("foo" recursion)+
+-----
+ERROR: This recursion never terminates
+HELP: A recursive expression must have a branch that doesn't reach the `recursion`, or can repeat 0 times
+SPAN: 7..16