Implement confusable detection.

crlf0710 · crlf0710 · commit 7786cb669c79 · 2020-04-25T14:17:52.000+08:00
diff --git a/Cargo.toml b/Cargo.toml
@@ -17,6 +17,7 @@ exclude = [ "target/*", "Cargo.lock" ]
 
 [dependencies]
 unicode-script = { version = "0.4.0", default-features = false }
+unicode-normalization = { version = "0.1.12", default-features = false }
 std = { version = "1.0", package = "rustc-std-workspace-std", optional = true }
 core = { version = "1.0", package = "rustc-std-workspace-core", optional = true }
 compiler_builtins = { version = "0.1", optional = true }
diff --git a/scripts/unicode.py b/scripts/unicode.py
@@ -81,6 +81,28 @@ def load_properties(f, interestingprops = None):
 
     return props
 
+def load_confusables(f):
+    fetch(f)
+    confusables = []
+    re1 = re.compile(r"^((?:[0-9A-F]+ )+);\t((?:[0-9A-F]+ )+);\t\w*")
+
+    for line in fileinput.input(os.path.basename(f), openhook=fileinput.hook_encoded("utf-8")):
+        d_input = 0
+        d_outputs = []
+        m = re1.match(line)
+        if not m:
+            continue
+        d_inputs = m.group(1).split()
+        if len(d_inputs) != 1:
+            raise Exception('More than one code point in first column')
+        d_input = int(d_inputs[0].strip(), 16)
+        for d_output in m.group(2).split():
+            d_outputitem = int(d_output, 16);
+            d_outputs.append(d_outputitem);
+        confusables.append((d_input, d_outputs))
+
+    return confusables
+
 def format_table_content(f, content, indent):
     line = " "*indent
     first = True
@@ -99,6 +121,18 @@ def format_table_content(f, content, indent):
 def escape_char(c):
     return "'\\u{%x}'" % c
 
+def escape_char_list(l):
+    line = "[";
+    first = True;
+    for c in l:
+        if first:
+            line += escape_char(c);
+        else:
+            line += ", " + escape_char(c);
+        first = False;
+    line += "]";
+    return line
+
 def emit_table(f, name, t_data, t_type = "&'static [(char, char)]", is_pub=True,
         pfun=lambda x: "(%s,%s)" % (escape_char(x[0]), escape_char(x[1])), is_const=True):
     pub_string = "const"
@@ -173,10 +207,51 @@ def emit_identifier_module(f):
             pfun=lambda x: "(%s,%s, IdentifierType::%s)" % (escape_char(x[0]), escape_char(x[1]), x[2]))
     f.write("}\n\n")
 
+def emit_confusable_detection_module(f):
+    f.write("pub mod confusable_detection {")
+    f.write("""
+
+    #[inline]
+    pub fn char_confusable_prototype(c: char) -> Option<&'static [char]> {
+        // FIXME: do we want to special case ASCII here?
+        match c as usize {
+            _ => super::util::bsearch_value_table(c, CONFUSABLES)
+        }
+    }
+
+""")
+
+    f.write("    // Confusable table:\n")
+    confusable_table = load_confusables("confusables.txt")
+    confusable_table.sort(key=lambda w: w[0])
+    
+    last_key = None
+    for (k, v) in confusable_table:
+        if k == last_key:
+            raise Exception("duplicate keys in confusables table: %s" % k)
+        last_key = k
+
+    emit_table(f, "CONFUSABLES", confusable_table, "&'static [(char, &'static [char])]", is_pub=False,
+            pfun=lambda x: "(%s, &%s)" % (escape_char(x[0]), escape_char_list(x[1])))
+    f.write("}\n\n")
+
+
 def emit_util_mod(f):
     f.write("""
 pub mod util {
     use core::result::Result::{Ok, Err};
+    
+    #[inline]
+    pub fn bsearch_value_table<T: Copy>(c: char, r: &'static [(char, T)]) -> Option<T> {
+        match r.binary_search_by_key(&c, |&(k, _)| k) {
+            Ok(idx) => {
+                let (_, v) = r[idx];
+                Some(v)
+            }
+            Err(_) => None
+        }
+    }
+    
     #[inline]
     pub fn bsearch_range_table(c: char, r: &'static [(char,char)]) -> bool {
         use core::cmp::Ordering::{Equal, Less, Greater};
@@ -224,3 +299,5 @@ def emit_util_mod(f):
         emit_util_mod(rf)
         ### identifier module
         emit_identifier_module(rf)
+        ### confusable_detection module
+        emit_confusable_detection_module(rf)
diff --git a/src/confusable_detection.rs b/src/confusable_detection.rs
@@ -0,0 +1,39 @@
+//! [Confusable detection](https://www.unicode.org/reports/tr39/#Confusable_Detection)
+
+use core::iter;
+
+enum OnceOrMore<T, I> {
+    Once(iter::Once<T>),
+    More(I),
+}
+
+impl<T, I> Iterator for OnceOrMore<T, I>
+where
+    I: Iterator<Item = T>,
+{
+    type Item = T;
+
+    fn next(&mut self) -> Option<T> {
+        use OnceOrMore::*;
+        match self {
+            Once(v) => v.next(),
+            More(i) => i.next(),
+        }
+    }
+}
+
+type StaticSliceIterCloned = core::iter::Cloned<core::slice::Iter<'static, char>>;
+
+fn char_prototype(c: char) -> OnceOrMore<char, StaticSliceIterCloned> {
+    use crate::tables::confusable_detection::char_confusable_prototype;
+    match char_confusable_prototype(c) {
+        None => OnceOrMore::Once(iter::once(c)),
+        Some(l) => OnceOrMore::More(l.iter().cloned()),
+    }
+}
+
+/// Calculate skeleton for string, as defined by UTS 39
+pub fn skeleton(s: &str) -> impl Iterator<Item = char> + '_ {
+    use unicode_normalization::UnicodeNormalization;
+    s.chars().nfd().flat_map(char_prototype).nfd()
+}
diff --git a/src/lib.rs b/src/lib.rs
@@ -58,10 +58,12 @@ extern crate test;
 
 pub use tables::UNICODE_VERSION;
 
+pub mod confusable_detection;
 pub mod general_security_profile;
 pub mod mixed_script;
 pub mod restriction_level;
 
+pub use confusable_detection::skeleton;
 pub use general_security_profile::GeneralSecurityProfile;
 pub use mixed_script::MixedScript;
 pub use restriction_level::{RestrictionLevel, RestrictionLevelDetection};
diff --git a/src/tables.rs b/src/tables.rs
diff --git a/src/tests.rs b/src/tests.rs
