Bring back the traces

components/segmenter/src/lib.rs

@@ -100,7 +100,7 @@
 //! See [`SentenceSegmenter`] for more examples.
 
 // https://github.com/unicode-org/icu4x/blob/main/docs/process/boilerplate.md#library-annotations
-#![cfg_attr(not(any(test, feature = "std")), no_std)]
+//#![cfg_attr(not(any(test, feature = "std")), no_std)]
 #![cfg_attr(
     not(test),
     deny(

components/segmenter/src/rule_segmenter.rs

@@ -105,6 +105,7 @@ impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> Iterator for RuleBreakIterator<'
             let left_prop = self.get_break_property(left_codepoint);
             self.advance_iter();
 
+            let right_codepoint = self.get_current_codepoint().map_or("????".to_string(), |c| format!("U+{:02X}", c.into()));
             let Some(right_prop) = self.get_current_break_property() else {
                 self.boundary_property = left_prop;
                 return Some(self.len);
@@ -126,14 +127,21 @@ impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> Iterator for RuleBreakIterator<'
 
             // If break_state is equals or grater than 0, it is alias of property.
             let mut break_state = self.get_break_state_from_table(left_prop, right_prop);
+            let STATE_NAMES = ["Unknown", "CR", "LF", "Extend", "Sep", "Format", "Sp", "Lower", "Upper", "OLetter", "Numeric", "ATerm", "SContinue", "STerm", "Close", "ATerm_Close", "ATerm_Close_Sp", "STerm_Close", "STerm_Close_Sp", "Upper_ATerm", "Lower_ATerm", "ATerm_Close_Sp_SB8", "ATerm_Close_Sp_ParaSep", "ATerm_Close_Sp_CR", "STerm_Close_Sp_ParaSep", "STerm_Close_Sp_CR", "sot", "eot"];
+            println!("left={:02X} right={:02X} {} state={:02X}", left_prop, right_prop, right_codepoint, break_state);
+            println!("left={} right={} {}", STATE_NAMES[left_prop as usize], STATE_NAMES[right_prop as usize], right_codepoint);
 
             if break_state >= 0 {
                 // This isn't simple rule set. We need marker to restore iterator to previous position.
                 let mut previous_iter = self.iter.clone();
                 let mut previous_pos_data = self.current_pos_data;
                 let mut previous_left_prop = left_prop;
 
+                if (break_state & INTERMEDIATE_MATCH_RULE) != 0 {
+                    println!("going through intermediate match rule");
+                }
                 break_state &= !INTERMEDIATE_MATCH_RULE;
+                println!("Inner loop");
                 loop {
                     self.advance_iter();
 
@@ -155,6 +163,8 @@ impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> Iterator for RuleBreakIterator<'
 
                     let previous_break_state = break_state;
                     break_state = self.get_break_state_from_table(break_state as u8, prop);
+                    println!("> left={:02X} right={:02X} state={:02X}", previous_break_state, prop, break_state);
+                    println!("> left={} right={}", STATE_NAMES[(previous_break_state & !INTERMEDIATE_MATCH_RULE) as usize], STATE_NAMES[prop as usize]);
                     if break_state < 0 {
                         break;
                     }
@@ -167,6 +177,7 @@ impl<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized> Iterator for RuleBreakIterator<'
                         previous_left_prop = break_state as u8;
                     }
                     if (break_state & INTERMEDIATE_MATCH_RULE) != 0 {
+                        println!("going through intermediate match rule");
                         break_state -= INTERMEDIATE_MATCH_RULE;
                         previous_iter = self.iter.clone();
                         previous_pos_data = self.current_pos_data;

components/segmenter/src/sentence.rs

@@ -28,6 +28,12 @@ pub struct SentenceBreakIterator<'l, 's, Y: RuleBreakType<'l, 's> + ?Sized>(
     RuleBreakIterator<'l, 's, Y>,
 );
 
+impl<'l, 's, Y: RuleBreakType<'l, 's>> SentenceBreakIterator<'l, 's, Y> {
+    pub fn state(&self) -> u8 {
+        self.0.boundary_property
+    }
+}
+
 derive_usize_iterator_with_type!(SentenceBreakIterator);
 
 /// Sentence break iterator for an `str` (a UTF-8 string).

components/segmenter/tests/spec_test.rs

@@ -223,6 +223,7 @@ fn sentence_break_test(filename: &str) {
             let mut iter = segmenter.segment_str(&s);
             // TODO(egg): It would be really nice to have Name here.
             println!("  | A | E | Code pt. | Sentence_Break | State | Literal");
+            let STATE_NAMES = ["Unknown", "CR", "LF", "Extend", "Sep", "Format", "Sp", "Lower", "Upper", "OLetter", "Numeric", "ATerm", "SContinue", "STerm", "Close", "ATerm_Close", "ATerm_Close_Sp", "STerm_Close", "STerm_Close_Sp", "Upper_ATerm", "Lower_ATerm", "ATerm_Close_Sp_SB8", "ATerm_Close_Sp_ParaSep", "ATerm_Close_Sp_CR", "STerm_Close_Sp_ParaSep", "STerm_Close_Sp_CR", "sot", "eot"];
             for (i, c) in s.char_indices() {
                 let expected_break = test.break_result_utf8.contains(&i);
                 let actual_break = result.contains(&i);
@@ -244,7 +245,7 @@ fn sentence_break_test(filename: &str) {
                         .unwrap_or(&format!("{:?}", sb.get(c))),
                     // Placeholder for logging the state if exposed.
                     // Not "?????" to hide from clippy.
-                    "?".repeat(5),
+                    if actual_break { format!("{:02X} {}", iter.state(), STATE_NAMES[iter.state() as usize]) } else {"?".repeat(5)},
                     c
                 )
             }

provider/datagen/src/transform/segmenter/mod.rs

@@ -17,6 +17,7 @@ use icu_provider::datagen::IterableDataProvider;
 use icu_provider::prelude::*;
 use icu_segmenter::provider::*;
 use icu_segmenter::symbols::*;
+use itertools::Itertools;
 use std::fmt::Debug;
 use zerovec::ZeroVec;
 
@@ -469,6 +470,8 @@ impl crate::DatagenProvider {
         // sot and eot
         properties_names.push("sot".to_string());
         properties_names.push("eot".to_string());
+        println!("{:?}", properties_names);
+        println!("{}", properties_names.iter().enumerate().map(|(i, name)| format!("{:02X}={}", i, name)).join("\n"));
 
         let rule_size = properties_names.len() * properties_names.len();
         let mut break_state_table = vec![UNKNOWN_RULE; rule_size];