levenshtein/regex: move levenshtein, remove regex

This is where the Levenshtein automata originally lived. I originally
moved it out of the crate in order to reduce dependencies. But we can
instead just make it an optional feature.

Instead of moving fst-regex back into the crate as well, we instead
recommend using `regex-automata`. `regex-automata` is basically the
productionized version of "compile a regex to a DFA," which is exactly
what we want.

Cargo.toml

@@ -15,8 +15,9 @@ license = "Unlicense/MIT"
 edition = "2018"
 
 [features]
-mmap = ["memmap"]
 default = ["mmap"]
+mmap = ["memmap"]
+levenshtein = ["utf8-ranges"]
 
 [[bench]]
 name = "build"
@@ -33,11 +34,10 @@ bench = true
 [dependencies]
 byteorder = "1"
 memmap = { version = "0.6.0", optional = true }
+utf8-ranges = { version = "1", optional = true }
 
 [dev-dependencies]
 fnv = "1.0.5"
-fst-levenshtein = { version = "0.2", path = "fst-levenshtein" }
-fst-regex = { version = "0.2", path = "fst-regex" }
 lazy_static = "0.2.8"
 quickcheck = { version = "0.7", default-features = false }
 rand = "0.5"

fst-bin/Cargo.toml

@@ -24,11 +24,10 @@ bit-set = "0.4"
 chan = "0.1"
 csv = "1.1.3"
 docopt = "1.1"
-fst = { path = "..", version = "0.3" }
-fst-regex = { path = "../fst-regex", version = "0.2" }
-fst-levenshtein = { path = "../fst-levenshtein", version = "0.2" }
+fst = { path = "..", version = "0.3", features = ["levenshtein"] }
 lines = "0.0"
 num_cpus = "1.5"
+regex-automata = { version = "*", path = "/home/andrew/rust/regex-automata", features = ["fst1"] }
 serde = { version = "1.0.104", features = ["derive"] }
 tempdir = "0.3"
 

fst-bin/src/cmd/fuzzy.rs

@@ -1,8 +1,8 @@
 use std::io;
 
 use docopt::Docopt;
+use fst::levenshtein::Levenshtein;
 use fst::raw::Fst;
-use fst_levenshtein::Levenshtein;
 use serde::Deserialize;
 
 use crate::util;

fst-bin/src/cmd/grep.rs

@@ -2,7 +2,8 @@ use std::io;
 
 use docopt::Docopt;
 use fst::raw::Fst;
-use fst_regex::Regex;
+use regex_automata::dense;
+// use regex_automata::sparse::SparseDFA;
 use serde::Deserialize;
 
 use crate::util;
@@ -40,8 +41,20 @@ pub fn run(argv: Vec<String>) -> Result<(), Error> {
         .and_then(|d| d.argv(&argv).deserialize())
         .unwrap_or_else(|e| e.exit());
     let fst = unsafe { Fst::from_path(&args.arg_fst) }?;
-    let lev = Regex::new(&args.arg_regex)?;
-    let mut q = fst.search(&lev);
+    let dense_dfa = dense::Builder::new()
+        .anchored(true)
+        .byte_classes(false)
+        .premultiply(false)
+        .build(&args.arg_regex)?;
+    let dfa = match dense_dfa {
+        dense::DenseDFA::Standard(dfa) => dfa,
+        _ => unreachable!(),
+    };
+    // let dfa = match dense_dfa.to_sparse()? {
+    // SparseDFA::Standard(dfa) => dfa,
+    // _ => unreachable!(),
+    // };
+    let mut q = fst.search(&dfa);
     if let Some(ref start) = args.flag_start {
         q = q.ge(start);
     }

fst-levenshtein/Cargo.toml

@@ -1,9 +1,9 @@
 [package]
 name = "fst-levenshtein"
-version = "0.2.1"  #:version
+version = "0.3.0"  #:version
 authors = ["Andrew Gallant <jamslam@gmail.com>"]
 description = """
-Search finite state transducers with fuzzy queries using Levenshtein automata.
+DEPRECATED. Use 'fst' crate with 'levenshtein' feature instead.
 """
 documentation = "https://docs.rs/fst-levenshtein"
 homepage = "https://github.com/BurntSushi/fst"

fst-regex/Cargo.toml

@@ -1,9 +1,9 @@
 [package]
 name = "fst-regex"
-version = "0.2.2"  #:version
+version = "0.3.0"  #:version
 authors = ["Andrew Gallant <jamslam@gmail.com>"]
 description = """
-Search finite state transducers with regular expression.
+DEPRECATED. Use 'regex-automata' crate with 'fst1' feature instead.
 """
 documentation = "https://docs.rs/fst-regex"
 homepage = "https://github.com/BurntSushi/fst"