many: use Mu::Regex instead of std::regex

The former is PCRE-compatible, and faster than std::regex.

lib/message/mu-message.cc

@@ -289,14 +289,14 @@ extract_tags(const MimeMessage& mime_msg)
 	constexpr std::array<std::pair<const char*, char>, 3> tag_headers = {{
 		{"X-Label", ' '}, {"X-Keywords", ','}, {"Keywords", ' '}
 	}};
-	static const auto strip_rx{std::regex("^\\s+| +$|( )\\s+")};
 
 	std::vector<std::string> tags;
 	seq_for_each(tag_headers, [&](auto&& item) {
 		if (auto&& hdr = mime_msg.header(item.first); hdr) {
 			for (auto&& tagval : split(*hdr, item.second)) {
-				tags.emplace_back(
-					std::regex_replace(tagval, strip_rx, "$1"));
+				tagval.erase(0, tagval.find_first_not_of(' '));
+				tagval.erase(tagval.find_last_not_of(' ')+1);
+				tags.emplace_back(std::move(tagval));
 			}
 		}
 	});
@@ -631,10 +631,10 @@ fill_document(Message::Private& priv)
 
 	const auto path{doc.string_value(Field::Id::Path)};
 	const auto refs{mime_msg.references()};
-        const auto& raw_message_id = mime_msg.message_id();
-        const auto message_id = raw_message_id.has_value() && !raw_message_id->empty()
-            ? *raw_message_id
-            : fake_message_id(path);
+	const auto& raw_message_id = mime_msg.message_id();
+	const auto message_id = raw_message_id.has_value() && !raw_message_id->empty()
+	    ? *raw_message_id
+	    : fake_message_id(path);
 
 	process_message(mime_msg, path, priv);
 

lib/mu-contacts-cache.cc

@@ -25,10 +25,10 @@
 #include <sstream>
 #include <functional>
 #include <algorithm>
-#include <regex>
 #include <ctime>
 
 #include <utils/mu-utils.hh>
+#include <utils/mu-regex.hh>
 #include <glib.h>
 
 using namespace Mu;
@@ -59,8 +59,8 @@ struct ContactsCache::Private {
 	ContactUMap contacts_;
 	std::mutex  mtx_;
 
-	const StringVec               personal_plain_;
-	const std::vector<std::regex> personal_rx_;
+	const StringVec			personal_plain_;
+	const std::vector<Regex>	personal_rx_;
 
 	size_t dirty_;
 
@@ -76,8 +76,8 @@ private:
 		StringVec svec;
 		std::copy_if(personal.begin(),  personal.end(),
 			     std::back_inserter(svec), [&](auto&& p) {
-				     return p.size() < 2
-					     || p.at(0) != '/' || p.at(p.length() - 1) != '/';
+				     return p.size() < 2 || p.at(0) != '/' ||
+					     p.at(p.length() - 1) != '/';
 			     });
 		return svec;
 	}
@@ -89,18 +89,20 @@ private:
 	 *
 	 * @return
 	 */
-	std::vector<std::regex> make_personal_rx(const StringVec& personal) const {
-		std::vector<std::regex> rxvec;
+	std::vector<Regex> make_personal_rx(const StringVec& personal) const {
+		std::vector<Regex> rxvec;
 		for(auto&& p: personal) {
 			if (p.size() < 2 || p[0] != '/' || p[p.length()- 1] != '/')
 				continue;
 			// a regex pattern.
 			try {
 				const auto rxstr{p.substr(1, p.length() - 2)};
-				rxvec.emplace_back(std::regex(
-				    rxstr, std::regex::basic | std::regex::optimize |
-				    std::regex::icase));
-			} catch (const std::regex_error& rex) {
+				auto opts = static_cast<GRegexCompileFlags>(G_REGEX_OPTIMIZE|G_REGEX_CASELESS);
+				auto rx = Regex::make(rxstr, opts);
+				if (!rx)
+					throw rx.error();
+				rxvec.emplace_back(rx.value());
+			} catch (const Error& rex) {
 				g_warning("invalid personal address regexp '%s': %s",
 					  p.c_str(),
 					  rex.what());
@@ -328,8 +330,7 @@ ContactsCache::is_personal(const std::string& addr) const
 			return true;
 
 	for (auto&& rx : priv_->personal_rx_) {
-		std::smatch m; // perhaps cache addr in personal_plain_?
-		if (std::regex_match(addr, m, rx))
+		if (rx.matches(addr))
 			return true;
 	}
 

lib/mu-parser.cc

@@ -19,12 +19,12 @@
 #include "mu-parser.hh"
 
 #include <algorithm>
-#include <regex>
 #include <limits>
 
 #include "mu-tokenizer.hh"
 #include "utils/mu-utils.hh"
 #include "utils/mu-error.hh"
+#include "utils/mu-regex.hh"
 #include "message/mu-message.hh"
 
 using namespace Mu;
@@ -64,7 +64,7 @@ struct Parser::Private {
 	Private(const Store& store, Parser::Flags flags) : store_{store}, flags_{flags} {}
 
 	std::vector<std::string> process_regex(const std::string& field,
-					       const std::regex&  rx) const;
+					       const Regex&  rx) const;
 
 	Mu::Tree term_1(Mu::Tokens& tokens, WarningVec& warnings) const;
 	Mu::Tree term_2(Mu::Tokens& tokens, Node::Type& op, WarningVec& warnings) const;
@@ -194,7 +194,7 @@ process_range(const std::string& field_str,
 
 std::vector<std::string>
 Parser::Private::process_regex(const std::string& field_str,
-			       const std::regex& rx) const
+			       const Regex& rx) const
 {
 	const auto field_opt{field_from_name(field_str)};
 	if (!field_opt)
@@ -204,7 +204,7 @@ Parser::Private::process_regex(const std::string& field_str,
 	std::vector<std::string> terms;
 	store_.for_each_term(field_opt->id, [&](auto&& str) {
 		auto val{str.c_str() + 1}; // strip off the Xapian prefix.
-		if (std::regex_search(val, rx))
+		if (rx.matches(val))
 			terms.emplace_back(std::move(val));
 		return true;
 	});
@@ -263,9 +263,11 @@ Parser::Private::regex(const FieldInfoVec& fields,
 
 	try {
 		Tree tree(Node{Node::Type::OpOr});
-		const auto rx = std::regex(rxstr);
+		const auto rx = Regex::make(rxstr, G_REGEX_OPTIMIZE);
+		if (!rx)
+			throw rx.error();
 		for (const auto& field : fields) {
-			const auto terms = process_regex(field.field, rx);
+			const auto terms = process_regex(field.field, *rx);
 			for (const auto& term : terms) {
 				tree.add_child(Tree({Node::Type::ValueAtomic,
 							FieldValue{field.id, term}}));

lib/tests/bench-indexer.cc

@@ -21,10 +21,10 @@
 #include <thread>
 #include <vector>
 #include <iostream>
-#include <regex>
 #include <fstream>
 
 #include <utils/mu-utils.hh>
+#include <utils/mu-regex.hh>
 #include <mu-store.hh>
 #include "mu-maildir.hh"
 
@@ -390,13 +390,15 @@ The archives can be found at:
 https://lists.cs.wisc.edu/archive/htcondor-users/
 --===============0678627779074767862==--)";
 
-
 static std::string
-message(const std::regex& rx, size_t id)
+message(const Regex& rx, size_t id)
 {
 	char buf[16];
 	::snprintf(buf, sizeof(buf), "%zu", id);
-	return std::regex_replace(test_msg, rx, buf);
+
+	return to_string_gchar(
+		g_regex_replace(rx, test_msg, -1, 0, buf,
+				G_REGEX_MATCH_DEFAULT, {}));
 }
 
 struct TestData {
@@ -420,7 +422,7 @@ setup(const TestData& tdata)
 		auto res = maildir_mkdir(mdir);
 		g_assert(!!res);
 	}
-	const auto rx = std::regex("@ID@");
+	const auto rx = Regex::make("@ID@");
 	/* create messages */
 	for (size_t n = 0; n != tdata.num_messages; ++n) {
 		auto mpath = format("%s/maildir-%zu/cur/msg-%zu:2,S",
@@ -428,7 +430,7 @@ setup(const TestData& tdata)
 				    n % tdata.num_maildirs,
 				    n);
 		std::ofstream stream(mpath);
-		auto msg = message(rx, n);
+		auto msg = message(*rx, n);
 		stream.write(msg.c_str(), msg.size());
 		g_assert_true(stream.good());
 	}

lib/utils/mu-utils.cc

@@ -39,7 +39,6 @@
 #include <cinttypes>
 #include <charconv>
 #include <limits>
-#include <regex>
 
 #include <glib.h>
 #include <glib/gprintf.h>
@@ -261,15 +260,6 @@ Mu::split(const std::string& str, char sepa)
 	return vec;
 }
 
-std::vector<std::string>
-Mu::split(const std::string& str, const std::regex& sepa_rx)
-{
-	std::sregex_token_iterator it(str.begin(), str.end(), sepa_rx, -1);
-	std::sregex_token_iterator end;
-
-	return {it, end};
-}
-
 std::string
 Mu::join(const std::vector<std::string>& svec, const std::string& sepa)
 {

lib/utils/mu-utils.hh

@@ -33,7 +33,6 @@
 #include <type_traits>
 #include <algorithm>
 #include <numeric>
-#include <regex>
 
 #include "mu-utils-format.hh"
 #include "mu-option.hh"
@@ -97,16 +96,6 @@ std::vector<std::string> split(const std::string& str, const std::string& sepa);
  */
 std::vector<std::string> split(const std::string& str, char sepa);
 
-/**
- * Split a string in parts
- *
- * @param str a string
- * @param sepa the separator regex
- *
- * @return the parts.
- */
-std::vector<std::string> split(const std::string& str, const std::regex& sepa_rx);
-
 /**
  * Join the strings in svec into a string, separated by sepa
  *
@@ -173,8 +162,6 @@ bool locale_workaround();
  */
 bool timezone_available(const std::string& tz);
 
-
-
 /**
  * Well-known runtime paths
  *

lib/utils/tests/test-utils.cc

@@ -220,9 +220,6 @@ test_split()
 	// char sepa
 	assert_equal_svec(split("axbxc", 'x'), {"a", "b", "c"});
 	assert_equal_svec(split("axbxcx", 'x'), {"a", "b", "c", ""});
-
-	// rx sexp
-	assert_equal_svec(split("axbyc", std::regex("[xy]")), {"a", "b", "c"});
 }
 
 static void

mu/mu-cmd-extract.cc

@@ -21,12 +21,11 @@
 #include "mu-cmd.hh"
 #include "utils/mu-util.h"
 #include "utils/mu-utils.hh"
+#include "utils/mu-regex.hh"
 #include <message/mu-message.hh>
-#include <regex>
 
 using namespace Mu;
 
-
 static Result<void>
 save_part(const Message::Part& part, size_t idx, const Options& opts)
 {
@@ -73,12 +72,13 @@ save_parts(const std::string& path, const std::string& filename_rx,
 			else if (seq_some(opts.extract.parts,
 				     [&](auto&& num){return num==partnum;}))
 				return true;
-			else if (!filename_rx.empty() && part.raw_filename() &&
-				 std::regex_match(*part.raw_filename(),
-						  std::regex{filename_rx}))
-				return true;
-			else
-				return false;
+			else if (!filename_rx.empty() && part.raw_filename()) {
+				if (auto rx = Regex::make(filename_rx); !rx)
+					throw rx.error();
+				else if (rx->matches(*part.raw_filename()))
+					return true;
+			}
+			return false;
 		});
 
 		if (!do_extract)