Merge pull request #267 from InterplanetaryEngineer/master

Accelerate symbolReferencesInternal further and improve refreshDocument algorithm

build.zig

@@ -237,14 +237,12 @@ pub fn build(b: *std.build.Builder) !void {
     const test_step = b.step("test", "Run all the tests");
     test_step.dependOn(builder.getInstallStep());
 
-    var unit_tests = b.addTest("tests/unit_tests.zig");
+    var unit_tests = b.addTest("src/unit_tests.zig");
-    unit_tests.addPackage(.{ .name = "analysis", .path = "src/analysis.zig" });
-    unit_tests.addPackage(.{ .name = "types", .path = "src/types.zig" });
     unit_tests.setBuildMode(.Debug);
+    test_step.dependOn(&unit_tests.step);
 
     var session_tests = b.addTest("tests/sessions.zig");
     session_tests.addPackage(.{ .name = "header", .path = "src/header.zig" });
     session_tests.setBuildMode(.Debug);
-
     test_step.dependOn(&session_tests.step);
 }

src/document_store.zig

@@ -360,51 +360,38 @@ fn refreshDocument(self: *DocumentStore, handle: *Handle, zig_lib_path: ?[]const
     handle.document_scope.deinit(self.allocator);
     handle.document_scope = try analysis.makeDocumentScope(self.allocator, handle.tree);
     
-    // TODO: Better algorithm or data structure?
+    var new_imports = std.ArrayList([]const u8).init(self.allocator);
-    // Removing the imports is costly since they live in an array list
+    errdefer new_imports.deinit();
-    // Perhaps we should use an AutoHashMap([]const u8, {}) ?
+    try analysis.collectImports(&new_imports, handle.tree);
 
-    // Try to detect removed imports and decrement their counts.
+    // Convert to URIs
-    if (handle.import_uris.items.len == 0) return;
+    var i: usize = 0;
-
+    while (i < new_imports.items.len) {
-    var arena = std.heap.ArenaAllocator.init(self.allocator);
+        if (try self.uriFromImportStr(self.allocator, handle.*, new_imports.items[i])) |uri| {
-    defer arena.deinit();
+            // The raw import strings are owned by the document and do not need to be freed here.
-
+            new_imports.items[i] = uri;
-    var import_strs = std.ArrayList([]const u8).init(&arena.allocator);
+            i += 1;
-    try analysis.collectImports(&import_strs, handle.tree);
+        } else {
-
+            _ = new_imports.swapRemove(i);
-    const still_exist = try arena.allocator.alloc(bool, handle.import_uris.items.len);
-    for (still_exist) |*ex| {
-        ex.* = false;
-    }
-
-    const std_uri = try stdUriFromLibPath(&arena.allocator, zig_lib_path);
-    for (import_strs.items) |str| {
-        const uri = (try self.uriFromImportStr(&arena.allocator, handle.*, str)) orelse continue;
-
-        exists_loop: for (still_exist) |*does_still_exist, i| {
-            if (does_still_exist.*) continue;
-
-            if (std.mem.eql(u8, handle.import_uris.items[i], uri)) {
-                does_still_exist.* = true;
-                break :exists_loop;
         }
     }
+
+    const old_imports = handle.import_uris;
+    handle.import_uris = new_imports;
+    defer old_imports.deinit();
+
+    // Remove all old_imports that do not exist anymore
+    for (old_imports.items) |old| {
+        still_exists: {
+            for (new_imports.items) |new| {
+                if (std.mem.eql(u8, new, old)) {
+                    break :still_exists;
                 }
-
-    // Go through still_exist, remove the items that are false and decrement their handle counts.
-    var idx: usize = 0;
-    for (still_exist) |does_still_exist| {
-        if (does_still_exist) {
-            idx += 1;
-            continue;
             }
-
+            log.debug("Import removed: {s}", .{old});
-        log.debug("Import removed: {s}", .{handle.import_uris.items[idx]});
+            self.decrementCount(old);
-        const uri = handle.import_uris.orderedRemove(idx);
+            self.allocator.free(old);
-
+        }
-        self.decrementCount(uri);
-        self.allocator.free(uri);
     }
 }
 
@@ -513,18 +500,14 @@ pub fn uriFromImportStr(
         }
         return null;
     } else {
-        // Find relative uri
+        const base = handle.uri();
-        const path = try URI.parse(allocator, handle.uri());
+        var base_len = base.len;
-        defer allocator.free(path);
+        while (base[base_len - 1] != '/' and base_len > 0) { base_len -= 1; }
-
+        base_len -= 1;
-        const dir_path = std.fs.path.dirname(path) orelse "";
+        if (base_len <= 0) {
-        const import_path = try std.fs.path.resolve(allocator, &[_][]const u8{
+            return error.UriBadScheme;
-            dir_path, import_str,
+        }
-        });
+        return try URI.pathRelative(allocator, base[0..base_len], import_str);
-
-        defer allocator.free(import_path);
-
-        return try URI.fromPath(allocator, import_path);
     }
 }
 

src/main.zig

@@ -1213,8 +1213,6 @@ fn loadConfig(folder_path: []const u8) ?Config {
     };
     defer allocator.free(file_buf);
 
-    // TODO: Uh oh. Profile the actual build time impact
-    // of adding config options and consider alternatives (TOML?)
     @setEvalBranchQuota(2000);
     // TODO: Better errors? Doesn't seem like std.json can provide us positions or context.
     var config = std.json.parse(Config, &std.json.TokenStream.init(file_buf), std.json.ParseOptions{ .allocator = allocator }) catch |err| {

src/unit_tests.zig

@@ -1,5 +1,7 @@
-const analysis = @import("analysis");
+const analysis = @import("analysis.zig");
-const types = @import("types");
+const types = @import("types.zig");
+const offsets = @import("offsets.zig");
+const URI = @import("uri.zig");
 
 const std = @import("std");
 
@@ -30,11 +32,11 @@ fn testContext(comptime line: []const u8, comptime tag: anytype, comptime range:
 
     const doc = try makeUnnamedDocument(final_line);
     defer freeDocument(doc);
+    var arena = std.heap.ArenaAllocator.init(allocator);
+    defer arena.deinit();
 
-    const ctx = try analysis.documentPositionContext(allocator, doc, types.Position{
+    const p = try offsets.documentPosition(doc, .{ .line = 0, .character = @intCast(i64, cursor_idx) }, .utf8);
-        .line = 0,
+    const ctx = try analysis.documentPositionContext(&arena, doc, p);
-        .character = @intCast(i64, cursor_idx),
-    });
 
     if (std.meta.activeTag(ctx) != tag) {
         std.debug.warn("Expected tag {}, got {}\n", .{ tag, std.meta.activeTag(ctx) });
@@ -100,3 +102,15 @@ test "documentPositionContext" {
         "Se",
     );
 }
+
+
+test "pathRelative and escapes" {
+    const join1 = try URI.pathRelative(allocator, "file://project/zig", "/src/main+.zig");
+    defer allocator.free(join1);
+    std.testing.expectEqualStrings("file://project/zig/src/main%2B.zig", join1);
+
+    const join2 = try URI.pathRelative(allocator, "file://project/zig/wow", "../]src]/]main.zig");
+    defer allocator.free(join2);
+    std.testing.expectEqualStrings("file://project/zig/%5Dsrc%5D/%5Dmain.zig", join2);
+}
+

src/uri.zig

@@ -1,13 +1,24 @@
 const std = @import("std");
+const mem = std.mem;
 
 const reserved_chars = &[_]u8{
     '!', '#', '$', '%', '&', '\'',
     '(', ')', '*', '+', ',', ':',
     ';', '=', '?', '@', '[', ']',
 };
+const reserved_escapes = comptime blk: {
+    var escapes: [reserved_chars.len][3]u8
+        = [_][3]u8{[_]u8{undefined} ** 3} ** reserved_chars.len;
+
+    for (reserved_chars) |c, i| {
+        escapes[i][0] = '%';
+        _ = std.fmt.bufPrint(escapes[i][1..], "{X}", .{c}) catch unreachable;
+    }
+    break :blk &escapes;
+};
 
 /// Returns a URI from a path, caller owns the memory allocated with `allocator`
-pub fn fromPath(allocator: *std.mem.Allocator, path: []const u8) ![]const u8 {
+pub fn fromPath(allocator: *mem.Allocator, path: []const u8) ![]const u8 {
     if (path.len == 0) return "";
     const prefix = if (std.builtin.os.tag == .windows) "file:///" else "file://";
 
@@ -19,10 +30,8 @@ pub fn fromPath(allocator: *std.mem.Allocator, path: []const u8) ![]const u8 {
     for (path) |char| {
         if (char == std.fs.path.sep) {
             try buf.append('/');
-        } else if (std.mem.indexOfScalar(u8, reserved_chars, char) != null) {
+        } else if (mem.indexOfScalar(u8, reserved_chars, char)) |reserved| {
-            // Write '%' + hex with uppercase
+            try buf.appendSlice(&reserved_escapes[reserved]);
-            try buf.append('%');
-            try std.fmt.format(out_stream, "{X}", .{char});
         } else {
             try buf.append(char);
         }
@@ -32,7 +41,7 @@ pub fn fromPath(allocator: *std.mem.Allocator, path: []const u8) ![]const u8 {
     if (std.builtin.os.tag == .windows) {
         if (buf.items.len > prefix.len + 1 and
             std.ascii.isAlpha(buf.items[prefix.len]) and
-            std.mem.startsWith(u8, buf.items[prefix.len + 1 ..], "%3A"))
+            mem.startsWith(u8, buf.items[prefix.len + 1 ..], "%3A"))
         {
             buf.items[prefix.len] = std.ascii.toLower(buf.items[prefix.len]);
         }
@@ -41,6 +50,47 @@ pub fn fromPath(allocator: *std.mem.Allocator, path: []const u8) ![]const u8 {
     return buf.toOwnedSlice();
 }
 
+/// Move along `rel` from `base` with a single allocation.
+/// `base` is a URI of a folder, `rel` is a raw relative path.
+pub fn pathRelative(allocator: *mem.Allocator, base: []const u8, rel: []const u8) ![]const u8 {
+    const max_size = base.len + rel.len * 3 + 1;
+
+    var result = try allocator.alloc(u8, max_size);
+    errdefer allocator.free(result);
+    mem.copy(u8, result, base);
+    var result_index: usize = base.len;
+
+    var it = mem.tokenize(rel, "/");
+    while (it.next()) |component| {
+        if (mem.eql(u8, component, ".")) {
+            continue;
+        } else if (mem.eql(u8, component, "..")) {
+            while (true) {
+                if (result_index == 0)
+                    return error.UriBadScheme;
+                result_index -= 1;
+                if (result[result_index] == '/')
+                    break;
+            }
+        } else {
+            result[result_index] = '/';
+            result_index += 1;
+            for (component) |char| {
+                if (mem.indexOfScalar(u8, reserved_chars, char)) |reserved| {
+                    const escape = &reserved_escapes[reserved];
+                    mem.copy(u8, result[result_index..], escape);
+                    result_index += escape.len;
+                } else {
+                    result[result_index] = char;
+                    result_index += 1;
+                }
+            }
+        }
+    }
+
+    return allocator.resize(result, result_index);
+}
+
 pub const UriParseError = error{
     UriBadScheme,
     UriBadHexChar,
@@ -59,8 +109,8 @@ fn parseHex(c: u8) !u8 {
 }
 
 /// Caller should free memory
-pub fn parse(allocator: *std.mem.Allocator, str: []const u8) ![]u8 {
+pub fn parse(allocator: *mem.Allocator, str: []const u8) ![]u8 {
-    if (str.len < 7 or !std.mem.eql(u8, "file://", str[0..7])) return error.UriBadScheme;
+    if (str.len < 7 or !mem.eql(u8, "file://", str[0..7])) return error.UriBadScheme;
 
     const uri = try allocator.alloc(u8, str.len - (if (std.fs.path.sep == '\\') 8 else 7));
     errdefer allocator.free(uri);
@@ -89,3 +139,4 @@ pub fn parse(allocator: *std.mem.Allocator, str: []const u8) ![]u8 {
 
     return allocator.shrink(uri, i);
 }
+