Merge branch 'main' of https://github.com/jsinger67/scnr

CHANGELOG.md

@@ -17,6 +17,13 @@ Be aware that this project is still v0.y.z which means that anything can change
 We defined for this project that while being on major version zero we mark incompatible changes with
 new minor version numbers. Please note that this is no version handling covered by `Semver`.
 
+## 0.7.2 - Not release yet
+
+- Refactor internal structs
+    - Refactored `CompiledNfa` to `CompiledDfa` for clarity
+    - Refactored `ScannerNfaImpl` to `ScannerImpl` for clarity
+- Minimization of the `CompiledDfa` for enhanced scanning performance
+
 ## 0.7.1 - 2025-01-21
 
 - Handle iterator exhaustion in FindMatchesImpl as partial fix for

README.md

@@ -54,7 +54,7 @@ fn main() {
 ## Guard rails
 
 * The scanners should be built quickly.
-* Scanners are based on NFAs internally.
+* Scanners are based on finite automata internally.
 * The scanners only support `&str`, i.e. patterns are of type `&str` and the input is of type
 `&str`. `scnr` focuses on programming languages rather than byte sequences.
 
@@ -114,8 +114,8 @@ The lookahead patterns denoted above as `S` are not considered as part of the ma
 
 ## Greediness of repetitions
 
-The generated scanners work with *compact NFAs* in which all repetition patterns like `*` and `+`
-match **greedily**.
+The generated scanners work with *compact DFAs* in which all repetition patterns like `*`, `+` and
+`?` match **greedily**.
 
 The `scnr` scanner generator does not directly support non-greedy quantifiers like *? or +? found in
 some other regex engines. However, you can achieve non-greedy behavior by carefully structuring your

doc/scnr.puml

@@ -145,7 +145,7 @@ package internal {
         - advance_char_indices_beyond_match(char_indices: &mut CharIndices, matched: Match)
 
     }
-    struct ScannerNfaImpl {
+    struct ScannerImpl {
         - current_mode: usize
         ~ match_char_class(char_class_id: CharClassID, c: char) -> bool
         ~ mode_name(&self, index: usize) -> Option<&str>
@@ -161,7 +161,7 @@ package internal {
     object SCANNER_CACHE<<(S, #FF7700) Singleton>>
     struct ScannerCacheEntry<<tuple>> {
         - modes: Vec<ScannerMode>
-        - nfa: ScannerNfaImpl
+        - scanner: ScannerImpl
     }
     struct CharacterClassRegistry {
     }
@@ -183,7 +183,7 @@ package internal {
     struct CompiledScannerMode {
         ~ name: String
     }
-    struct CompiledNfa {
+    struct CompiledDfa {
         - pattern: String
         - end_states: Vec<StateSetID>
     }
@@ -205,10 +205,10 @@ package internal {
         ~ terminal: TerminalID,
     }
 
-    FindMatchesImpl *--> ScannerNfaImpl: - scanner_impl
+    FindMatchesImpl *--> ScannerImpl: - scanner_impl
 
-    ScannerNfaImpl *--> CharacterClassRegistry: ~ character_classes
-    ScannerNfaImpl *--> "*" CompiledScannerMode: ~ scanner_modes
+    ScannerImpl *--> CharacterClassRegistry: ~ character_classes
+    ScannerImpl *--> "*" CompiledScannerMode: ~ scanner_modes
 
     CharacterClassRegistry *--> "*" CharacterClass: - character_classes
 
@@ -233,8 +233,8 @@ package internal {
     CompiledScannerMode *--> "*" NfaWithTerminal: ~ nfas
     CompiledScannerMode *--> "*" ScannerModeTransition: ~ transitions
 
-    CompiledNfa *--> "*" StateData: ~ states
-    CompiledNfa "Option" *--> CompiledLookahead: ~ lookahead
+    CompiledDfa *--> "*" StateData: ~ states
+    CompiledDfa "Option" *--> CompiledLookahead: ~ lookahead
     ScannerCache <-- SCANNER_CACHE: + instance_of
 
 }
@@ -255,7 +255,7 @@ ScannerBuilder *--> "1*" ScannerMode: - scanner_modes
 ScannerBuilder .> Scanner: build()
 ScannerBuilder ...> SCANNER_CACHE: uses
 
-Scanner *--> internal.ScannerNfaImpl: - inner
+Scanner *--> internal.ScannerImpl: - inner
 Scanner -|> ScannerModeSwitcher: implements
 Scanner .> FindMatches: find_iter()
 

src/find_matches.rs

@@ -1,7 +1,7 @@
 use log::trace;
 
 use crate::{
-    internal::{find_matches_impl::FindMatchesImpl, ScannerNfaImpl},
+    internal::{find_matches_impl::FindMatchesImpl, ScannerImpl},
     Match, Position, PositionProvider, ScannerModeSwitcher,
 };
 
@@ -34,7 +34,7 @@ pub struct FindMatches<'h> {
 
 impl<'h> FindMatches<'h> {
     /// Creates a new `FindMatches` iterator.
-    pub(crate) fn new(scanner_impl: ScannerNfaImpl, input: &'h str) -> Self {
+    pub(crate) fn new(scanner_impl: ScannerImpl, input: &'h str) -> Self {
         Self {
             inner: FindMatchesImpl::new(scanner_impl, input),
         }

src/internal/character_class_registry.rs

@@ -62,6 +62,12 @@ impl CharacterClassRegistry {
         self.character_classes.is_empty()
     }
 
+    /// Creates a match function for the character classes in the registry.
+    ///
+    /// Safety:
+    ///     The callers ensure that the character classes in the registry are valid.
+    ///     All character classes in the registry are valid which is guaranteed by the construction
+    ///     of the registry.
     pub(crate) fn create_match_char_class(
         &self,
     ) -> Result<Box<dyn (Fn(CharClassID, char) -> bool) + 'static + Send + Sync>> {
@@ -76,7 +82,7 @@ impl CharacterClassRegistry {
                 })?;
         Ok(Box::new(move |char_class, c| {
             // trace!("Match char class #{} '{}' -> {:?}", char_class.id(), c, res);
-            match_functions[char_class].call(c)
+            unsafe { match_functions.get_unchecked(char_class.as_usize()).call(c) }
         }))
     }
 }