Merge pull request #3649 from lightpanda-io/cli-suggest-threshold

cli: scale the "did you mean" distance with the input's length
This commit is contained in:
Karl Seguin authored and GitHub committed 2026-09-28 16:05:53 +08:00
commit cfad479548
5 files changed
+114 -45

No files matched your search

+42 -13
View File
@@ -69,8 +69,7 @@ fn logFilterValidator(allocator: Allocator, args: *std.process.Args.Iterator, li
}
const v = std.meta.stringToEnum(log.Scope, name) orelse {
log.fatal(.app, "invalid option choice", .{ .arg = "--log-filter", .value = part });
return error.InvalidOption;
return cli.invalidChoice("--log-filter", part[0 .. part.len - name.len], name, comptime tagNames(log.Scope) ++ &[_][]const u8{"all"});
};
try list.append(allocator, .{ .scope = v, .enable = enable });
@@ -85,10 +84,7 @@ fn logLevelValidator(_: Allocator, args: *std.process.Args.Iterator, target: *?l
return;
}
target.* = std.meta.stringToEnum(log.Level, str) orelse {
log.fatal(.app, "invalid option choice", .{ .arg = "--log-level", .value = str });
return error.InvalidArgument;
};
target.* = std.meta.stringToEnum(log.Level, str) orelse return cli.invalidChoice("--log-level", "", str, tagNames(log.Level));
log.opts.level = target.*.?;
}
@@ -101,10 +97,7 @@ fn mcpLogDefaults() void {
fn logFormatValidator(_: Allocator, args: *std.process.Args.Iterator, target: *?log.Format) !void {
const str = args.next() orelse return error.MissingArgument;
const format = std.meta.stringToEnum(log.Format, str) orelse {
log.fatal(.app, "invalid option choice", .{ .arg = "--log-format", .value = str });
return error.InvalidArgument;
};
const format = std.meta.stringToEnum(log.Format, str) orelse return cli.invalidChoice("--log-format", "", str, tagNames(log.Format));
target.* = format;
log.opts.format = format;
}
@@ -323,9 +316,8 @@ fn dumpValidator(_: Allocator, args: *std.process.Args.Iterator, target: *?DumpF
if (peek_args.next()) |next_arg| {
const mode = std.meta.stringToEnum(DumpFormat, next_arg) orelse {
// Anything else is the positional url, unless it is a misspelt format.
if (string.closest(next_arg, tagNames(DumpFormat), 2)) |near| {
log.fatal(.app, "invalid option choice", .{ .arg = "--dump", .value = log.red(next_arg), .did_you_mean = log.green(near) });
return error.InvalidArgument;
if (!cli.isUrlLike(next_arg) and string.closest(next_arg, tagNames(DumpFormat)) != null) {
return cli.invalidChoice("--dump", "", next_arg, tagNames(DumpFormat));
}
target.* = .html;
return;
@@ -1368,6 +1360,43 @@ test "Config: parseArgs --http-session-timeout" {
}
}
test "Config: parseArgs --dump tells a url from a misspelt format" {
var arena = std.heap.ArenaAllocator.init(std.testing.allocator);
defer arena.deinit();
{
const argv = [_][*:0]const u8{ "lightpanda", "fetch", "--dump", "markdown.com" };
const proc_args: std.process.Args = .{ .vector = &argv };
const config = try parseArgs(arena.allocator(), proc_args);
try std.testing.expectEqual(.html, config.mode.fetch.dump);
try std.testing.expectEqualStrings("markdown.com", config.mode.fetch.url.items[0]);
}
{
log.expectLog(&.{.app});
const argv = [_][*:0]const u8{ "lightpanda", "fetch", "--dump", "markdwon", "https://example.com" };
const proc_args: std.process.Args = .{ .vector = &argv };
try std.testing.expectError(error.InvalidArgument, parseArgs(arena.allocator(), proc_args));
}
}
test "Config: parseArgs tells a url from a misspelt command" {
var arena = std.heap.ArenaAllocator.init(std.testing.allocator);
defer arena.deinit();
{
const argv = [_][*:0]const u8{ "lightpanda", "version.io" };
const proc_args: std.process.Args = .{ .vector = &argv };
const config = try parseArgs(arena.allocator(), proc_args);
try std.testing.expectEqualStrings("version.io", config.mode.fetch.url.items[0]);
}
{
log.expectLog(&.{.app});
const argv = [_][*:0]const u8{ "lightpanda", "versoin" };
const proc_args: std.process.Args = .{ .vector = &argv };
try std.testing.expectError(error.UnknownCommand, parseArgs(arena.allocator(), proc_args));
}
}
test "Config: validateUserAgent" {
try validateUserAgent("Lightpanda/1.0");
try std.testing.expectError(error.Reserved, validateUserAgent("mozilla/1.0"));
+2 -2
View File
@@ -103,7 +103,7 @@ const meta_names = blk: {
/// Every slash-invocable name: browser tools, LLM triggers, meta commands.
pub const all_names = browser_tools.names ++ llm_names ++ meta_names;
/// Closest command name within two edits, or null — for "did you mean?" on typos.
/// Closest command name, or null — for "did you mean?" on typos.
pub fn closestCommand(name: []const u8) ?[]const u8 {
return string.closest(name, &all_names, 2);
return string.closest(name, &all_names);
}
+24 -12
View File
@@ -200,6 +200,26 @@ pub fn tagNames(comptime E: type) []const []const u8 {
};
}
/// No command or choice has a `.`, `/` or `:`, so `markdown.com` is a url
/// however close it is to `markdown`.
pub fn isUrlLike(arg: []const u8) bool {
return std.mem.indexOfAny(u8, arg, ".:/") != null;
}
/// `prefix` was stripped from `value` before matching, like `--log-filter`'s sign.
pub fn invalidChoice(arg: []const u8, prefix: []const u8, value: []const u8, choices: []const []const u8) error{InvalidArgument} {
var value_buf: [128]u8 = undefined;
const typed = std.fmt.bufPrint(&value_buf, "{s}{s}", .{ prefix, value }) catch value;
if (string.closest(value, choices)) |near| {
var near_buf: [128]u8 = undefined;
const suggestion = std.fmt.bufPrint(&near_buf, "{s}{s}", .{ prefix, near }) catch near;
log.fatal(.app, "invalid option choice", .{ .arg = arg, .value = log.red(typed), .did_you_mean = log.green(suggestion) });
} else {
log.fatal(.app, "invalid option choice", .{ .arg = arg, .value = log.red(typed) });
}
return error.InvalidArgument;
}
pub fn Builder(comptime commands: anytype) type {
return struct {
const Self = @This();
@@ -442,7 +462,7 @@ pub fn Builder(comptime commands: anytype) type {
}
// A bare word close to a command name is a typo, not a fetch url.
if (std.mem.startsWith(u8, cmd_str, "--") == false and string.closest(cmd_str, command_names, 2) != null) {
if (std.mem.startsWith(u8, cmd_str, "--") == false and !isUrlLike(cmd_str) and string.closest(cmd_str, command_names) != null) {
return unknownCommand(cmd_str);
}
@@ -474,7 +494,7 @@ pub fn Builder(comptime commands: anytype) type {
fn unknownCommand(name: []const u8) error{UnknownCommand} {
const arg = log.red(name);
if (string.closest(name, command_names, 2)) |near| {
if (string.closest(name, command_names)) |near| {
log.fatal(.app, "unknown command", .{ .arg = arg, .did_you_mean = log.green(near) });
} else {
log.fatal(.app, "unknown command", .{ .arg = arg });
@@ -688,15 +708,7 @@ pub fn Builder(comptime commands: anytype) type {
};
const str = args.next() orelse return error.MissingArgument;
const v = std.meta.stringToEnum(E, str) orelse {
const value = log.red(str);
if (string.closest(str, tagNames(E), 2)) |near| {
log.fatal(.app, "invalid option choice", .{ .arg = kebab_cased, .value = value, .did_you_mean = log.green(near) });
} else {
log.fatal(.app, "invalid option choice", .{ .arg = kebab_cased, .value = value });
}
return error.InvalidArgument;
};
const v = std.meta.stringToEnum(E, str) orelse return invalidChoice(kebab_cased, "", str, tagNames(E));
if (is_multiple) {
try target.append(allocator, v);
@@ -819,7 +831,7 @@ pub fn Builder(comptime commands: anytype) type {
if (std.mem.startsWith(u8, option_name, "--")) {
const names = comptime optionNames(options) ++ &[_][]const u8{"--help"};
const arg = log.red(option_name);
if (string.closest(option_name, names, 2)) |near| {
if (string.closest(option_name, names)) |near| {
log.fatal(.app, "unknown argument", .{ .mode = command.name, .arg = arg, .did_you_mean = log.green(near) });
} else {
log.fatal(.app, "unknown argument", .{ .mode = command.name, .arg = arg });
-1
View File
@@ -68,7 +68,6 @@ fn run(allocator: Allocator, main_arena: Allocator, proc_args: std.process.Args)
// Already logged where they were found.
error.UnknownCommand,
error.UnknownOption,
error.InvalidOption,
error.InvalidArgument,
error.MissingArgument,
error.TooManyPositionalArguments,
+46 -17
View File
@@ -417,28 +417,46 @@ pub fn isOneOf(needle: []const u8, haystack: []const []const u8) bool {
} else false;
}
/// Case-insensitive. Inputs over 64 bytes return `maxInt`; that fits the
/// longest CLI flag.
/// Case-insensitive, and swapping two adjacent characters counts as one
/// edit. Inputs over 64 bytes return `maxInt`; that fits the longest CLI flag.
fn editDistance(a: []const u8, b: []const u8) usize {
const max = 64;
if (a.len > max or b.len > max) return std.math.maxInt(usize);
var a_buf: [max]u8 = undefined;
var b_buf: [max]u8 = undefined;
const la = std.ascii.lowerString(&a_buf, a);
const lb = std.ascii.lowerString(&b_buf, b);
var prev2: [max + 1]u8 = undefined;
var prev: [max + 1]u8 = undefined;
var cur: [max + 1]u8 = undefined;
for (0..b.len + 1) |j| prev[j] = @intCast(j);
for (a, 1..) |ca, i| {
const la = std.ascii.toLower(ca);
for (0..lb.len + 1) |j| prev[j] = @intCast(j);
for (la, 1..) |ca, i| {
cur[0] = @intCast(i);
for (b, 1..) |cb, j| {
const cost: u8 = if (la == std.ascii.toLower(cb)) 0 else 1;
for (lb, 1..) |cb, j| {
const cost: u8 = if (ca == cb) 0 else 1;
cur[j] = @min(@min(prev[j] + 1, cur[j - 1] + 1), prev[j - 1] + cost);
if (i > 1 and j > 1 and ca == lb[j - 2] and la[i - 2] == cb) {
cur[j] = @min(cur[j], prev2[j - 2] + 1);
}
}
prev2 = prev;
prev = cur;
}
return prev[b.len];
}
/// Earlier candidates win ties.
pub fn closest(name: []const u8, candidates: []const []const u8, max_dist: usize) ?[]const u8 {
/// The candidate nearest to `name`, within about one edit per three characters
/// (rustc's rule). A prefix every candidate shares, like a flag's `--`, doesn't
/// count toward the length. Earlier candidates win ties.
pub fn closest(name: []const u8, candidates: []const []const u8) ?[]const u8 {
if (candidates.len == 0) return null;
var shared = candidates[0];
for (candidates[1..]) |cand| {
shared = shared[0 .. std.mem.indexOfDiff(u8, shared, cand) orelse shared.len];
}
const typed = if (std.mem.startsWith(u8, name, shared)) name.len - shared.len else name.len;
const max_dist = @max(typed, 3) / 3;
var best: ?[]const u8 = null;
var best_dist: usize = std.math.maxInt(usize);
for (candidates) |cand| {
@@ -553,6 +571,8 @@ test "editDistance" {
try testing.expectEqual(@as(usize, 1), editDistance("wait-m", "wait-ms"));
try testing.expectEqual(@as(usize, 1), editDistance("wait_ms", "wait-ms"));
try testing.expectEqual(@as(usize, 3), editDistance("kitten", "sitting"));
try testing.expectEqual(@as(usize, 1), editDistance("dmup", "dump"));
try testing.expectEqual(@as(usize, 1), editDistance("ab", "ba"));
try testing.expectEqual(@as(usize, 3), editDistance("", "abc"));
try testing.expectEqual(@as(usize, 3), editDistance("abc", ""));
@@ -562,15 +582,24 @@ test "editDistance" {
}
test "closest" {
const names = [_][]const u8{ "--dump", "--wait-ms", "--wait-until" };
try testing.expectEqual("--wait-ms", closest("--wait-mss", &names, 2));
try testing.expectEqual("--dump", closest("--dmup", &names, 2));
try testing.expectEqual(null, closest("--totally-wrong", &names, 2));
try testing.expectEqual(null, closest("--wait-mss", &names, 0));
try testing.expectEqual(null, closest("--dump", &.{}, 2));
const names = [_][]const u8{ "--dump", "--wait-ms", "--wait-until", "--insecure-disable-tls-host-verification" };
try testing.expectEqual("--wait-ms", closest("--wait-mss", &names));
try testing.expectEqual("--dump", closest("--dmup", &names));
try testing.expectEqual(null, closest("--dmpx", &names));
try testing.expectEqual(null, closest("--totally-wrong", &names));
try testing.expectEqual(null, closest("--dump", &.{}));
try testing.expectEqual("--insecure-disable-tls-host-verification", closest("--insecure-disable-tls-verification", &names));
const tie = [_][]const u8{ "ab", "ac" };
try testing.expectEqual("ab", closest("a", &tie, 1));
const formats = [_][]const u8{ "html", "markdown", "pdf", "png" };
try testing.expectEqual(null, closest("md", &formats));
try testing.expectEqual("pdf", closest("pdg", &formats));
const commands = [_][]const u8{ "fetch", "mcp", "run" };
try testing.expectEqual("run", closest("fun", &commands));
try testing.expectEqual(null, closest("ab", &commands));
const tie = [_][]const u8{ "abcd", "abce" };
try testing.expectEqual("abcd", closest("abc", &tie));
}
test "latin1ToUtf8" {