diff options
| author | Gabriel Schneider <[email protected]> | 2026-11-17 00:41:40 -0300 |
|---|---|---|
| committer | Gabriel Schneider <[email protected]> | 2026-11-17 12:15:01 -0300 |
| commit | 65612a1cc13dd2a7e866f88176dbf6db5dd83467 (patch) | |
| tree | d2b6173b9650532ff8fa8b489ca7506b75d93cbd /src/Plumb.zig | |
| parent | 797e1b05324a8de7217b7e483496f44c1e7bcdbd (diff) | |
| download | pardes-65612a1cc13dd2a7e866f88176dbf6db5dd83467.tar.gz pardes-65612a1cc13dd2a7e866f88176dbf6db5dd83467.zip | |
plumber: Plan 9's plumber in the core, its language and its files (docs/next-plumber.md, phase 1): src/Plumb.zig owns the rules, kept as their text (Core.plumber, digested and dumped; `include basic`'s, the embedded src/plumbing, until any are written, so a session and a dump from before it start with the defaults) and parsed again for each message, and a matcher over them read from rules.c and match.c: the message's pack and unpack (plumbunpackpartial's whole, short of its data, malformed), packed attributes read and quoted rc's way, rule sets of `object verb arg` lines, variables, include (basic, the configuration directory's, an absolute path, through the taped read site), rc's quoting and `$` expansion at parse and at match, every parse refusal in rules.c's words with its file and line, `$1`-`$9` refused in words (the regular expressions have no capture groups), `matches` over regexp.zig whole (`^(...)$`) or around a click counted in runes with Pike's two patterns, `is`, `set` rewriting for good, `add`, `delete` failing on an attribute not there as match.c's code does, `isfile` and `isdir` finding a name as Look does (past its place, the message's directory, a look's jump list's, Incl's) at the taped resolve sites, once a send; dispose to a port pardes reads (edit: opened as a look opens it, at addr=; web: the host's link opener; a loaded kind by its name), to a rule set's start or client run as exec runs a line (a builtin in place, else a command pane in the message's wdir: Core.exec_dir, exec.runCommand's reuse skipping the sender), or Plan 9's words for what went nowhere; rules print back in the order they were written, a variable where it was set, so a printed text reads back the same; plumb/rules (a write adds rule sets as each is whole, the rest at the close, `>` replaces them, one writer at a time) and plumb/send (a message put together across writes) in the tree, with fs-help's line, the reference's section and tree, and the fstree test's paths; nothing calls the plumber from Look yet; test: Plumb.zig's eleven (every parse error, print and read back, plan9port's basic less its submatch rule sets, its ¡- classes UTF-8's bytes, whole and click matches with horse.gift, quoting, a set's rewrite kept, delete of an absent attribute, dst filtering, isfile and isdir, dispose, a dump keeping the rules), test/behaviour/plumber.zig's two through the tree (Session records the link a host would open); the root's listings in tree.zig's and the process tests gain plumb, and the README folds a blank line to stay 45 lines; the changelog says so
Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_016Y9MNBECMWcJD5pdS4SR6T
Diffstat (limited to 'src/Plumb.zig')
| -rw-r--r-- | src/Plumb.zig | 1241 |
1 files changed, 1241 insertions, 0 deletions
diff --git a/src/Plumb.zig b/src/Plumb.zig new file mode 100644 index 00000000..57a19b14 --- /dev/null +++ b/src/Plumb.zig @@ -0,0 +1,1241 @@ +//! The plumber (Plan 9's plumber(4), windows/plumb/): the rules, and a +//! message matched against them and sent where they say. The language is +//! plumb(6)'s as rules.c and match.c read it, and its refusals are their +//! words; docs/next-plumber.md is the design. The rules are kept as their +//! text, a step's state as any other (digested, dumped, replicated), and +//! parsed again for each message: a send reads no file, and asks the disk +//! only at `isfile` and `isdir`, through the taped resolve sites Look uses. +const Plumb = @This(); +const std = @import("std"); +const pardes = @import("pardes.zig"); +const fs = @import("fs.zig"); +const look = @import("look.zig"); +const exec = @import("exec.zig"); +const layout = @import("layout.zig"); +const Kind = @import("Kind.zig"); +const Regex = @import("regexp.zig").Regex; +const Pardes = pardes.Pardes; + +/// The rules in force once any were written, as plumb/rules reads them: +/// what `print` makes of them, an include's text in its place, in the +/// order they were written. +text: std.ArrayList(u8) = .empty, +/// Whether rules were written (or cleared): until then the rules in force +/// are `include basic`'s, so a session starts with the defaults, and so +/// does a dump made before the plumber. +written: bool = false, + +/// The default rules, which `include basic` names, as Plan 9's names +/// /lib/plumb/basic. +pub const basic = @embedFile("plumbing"); + +/// The most rule text kept (Plan 9's whole basic is about 4 KiB), variables +/// and ports, as Plan 9's NDIR less its fixed files; deeper includes, as +/// rules.c's pushinput. +pub const max_text = 64 * 1024; +const max_vars = 32; +const max_ports = 32; +const max_depth = 10; + +/// fsys.c's and match.c's words for a message that went nowhere. +pub const e_bad_msg = "bad plumb message format"; +pub const e_no_port = "couldn't find destination for message"; +pub const e_no_start = "no start action for plumb message"; +pub const e_client = "no port for \"client\" rule"; +pub const e_empty = "empty argument list"; + +pub fn deinit(pl: *Plumb, gpa: std.mem.Allocator) void { + pl.text.deinit(gpa); +} + +// ---- the message ---- + +/// A plumb message (plumb.h, plumb(6)): six lines of header, then `data`. +pub const Msg = struct { + src: []const u8 = "", + dst: []const u8 = "", + wdir: []const u8 = "", + type: []const u8 = "", + /// `name=value` pairs, blank-separated, as they are packed: a value + /// holding a blank, a quote, `=` or a tab quoted rc's way. + attr: []const u8 = "", + data: []const u8 = "", + + /// plumbpack: the header lines, the data's length, the data. + pub fn pack(m: Msg, w: *std.Io.Writer) std.Io.Writer.Error!void { + try w.print("{s}\n{s}\n{s}\n{s}\n{s}\n{d}\n{s}", .{ m.src, m.dst, m.wdir, m.type, m.attr, m.data.len, m.data }); + } +}; + +/// plumbunpackpartial: a whole message, `more` bytes of its data still to +/// come, or one malformed (a header line short, a count the data passes). +pub const Unpacked = union(enum) { msg: Msg, more: usize, bad }; + +pub fn unpack(bytes: []const u8) Unpacked { + var lines: [6][]const u8 = undefined; + var at: usize = 0; + for (&lines) |*line| { + const nl = std.mem.indexOfScalarPos(u8, bytes, at, '\n') orelse return .bad; + line.* = bytes[at..nl]; + at = nl + 1; + } + // atoi's: digits after blanks, else 0. + const count = std.mem.trimStart(u8, lines[5], " \t"); + var n: usize = 0; + for (count) |c| { + if (c < '0' or c > '9') break; + n = std.math.mul(usize, n, 10) catch return .bad; + n = std.math.add(usize, n, c - '0') catch return .bad; + } + const rest = bytes.len - at; + if (n > rest) return .{ .more = n - rest }; + if (n < rest) return .bad; + return .{ .msg = .{ .src = lines[0], .dst = lines[1], .wdir = lines[2], .type = lines[3], .attr = lines[4], .data = bytes[at..] } }; +} + +/// One `name=value` of packed attributes, its value unquoted. +const Attr = struct { name: []const u8, value: []const u8 }; + +/// The attribute at `at.*` in `attr`, read as plumbunpackattr reads it, +/// its value unquoted into `buf`; null at their end, or at one with no +/// `=`, where Plan 9 stops. +fn nextAttr(attr: []const u8, at: *usize, buf: *[4096]u8) ?Attr { + var i = at.*; + while (i < attr.len and (attr[i] == ' ' or attr[i] == '\t')) i += 1; + if (i >= attr.len or attr[i] == '\n') return null; + const start = i; + while (i < attr.len and attr[i] != '=' and attr[i] != ' ' and attr[i] != '\t' and attr[i] != '\n') i += 1; + if (i >= attr.len or attr[i] != '=') return null; + const name = attr[start..i]; + i += 1; + var n: usize = 0; + var quoting = false; + while (i < attr.len and attr[i] != '\n' and n < buf.len) { + const c = attr[i]; + i += 1; + if (c == '\'') { + if (!quoting) { + quoting = true; + continue; + } + if (i < attr.len and attr[i] == '\'') { + i += 1; + } else { + quoting = false; + continue; + } + } else if (!quoting and (c == ' ' or c == '\t')) break; + buf[n] = c; + n += 1; + } + at.* = i; + return .{ .name = name, .value = buf[0..n] }; +} + +/// plumblookup: the first attribute named `name`'s value. +fn attrValue(attr: []const u8, name: []const u8, buf: *[4096]u8) ?[]const u8 { + var at: usize = 0; + while (nextAttr(attr, &at, buf)) |a| if (std.mem.eql(u8, a.name, name)) return a.value; + return null; +} + +/// `attr` packed again (plumbpackattr): the first attribute named `drop` +/// left out (plumbdelattr), those of `add` after it (plumbaddattr). +fn repack(arena: std.mem.Allocator, attr: []const u8, drop: ?[]const u8, add: []const u8) error{OutOfMemory}![]const u8 { + var out: std.Io.Writer.Allocating = .init(arena); + var dropped = drop == null; + var buf: [4096]u8 = undefined; + for ([2][]const u8{ attr, add }) |from| { + var at: usize = 0; + while (nextAttr(from, &at, &buf)) |a| { + if (!dropped and std.mem.eql(u8, a.name, drop.?)) { + dropped = true; + continue; + } + const w = &out.writer; + if (out.written().len > 0) w.writeByte(' ') catch return error.OutOfMemory; + w.print("{s}=", .{a.name}) catch return error.OutOfMemory; + // mesg.c's quote: a value with a blank, a quote, `=` or a tab. + if (std.mem.indexOfAny(u8, a.value, " '=\t") == null) { + w.writeAll(a.value) catch return error.OutOfMemory; + continue; + } + w.writeByte('\'') catch return error.OutOfMemory; + for (a.value) |c| { + if (c == '\'') w.writeByte('\'') catch return error.OutOfMemory; + w.writeByte(c) catch return error.OutOfMemory; + } + w.writeByte('\'') catch return error.OutOfMemory; + } + } + return out.written(); +} + +// ---- the rules ---- + +/// rules.c's objects and verbs, in its order (`kind` is an old `type`). +const Obj = enum { arg, attr, data, dst, plumb, src, type, wdir }; +const Verb = enum { add, client, delete, is, isdir, isfile, matches, set, start, to }; + +/// A line `object verb arg`: `qarg` is `arg` expanded with the variables +/// alone, its first word, as rules.c's parserule keeps it. +const Rule = struct { obj: Obj, verb: Verb, arg: []const u8, qarg: []const u8 }; +const Set = struct { pats: []const Rule, acts: []const Rule, port: ?[]const u8 }; +const Var = struct { name: []const u8, value: []const u8, qvalue: []const u8 }; + +/// The rules as read, each item where it was written: printed back in +/// that order, a variable given again later means what it meant to the +/// rules between (Plan 9's printrules puts every variable first). +const Rules = struct { + items: std.ArrayList(union(enum) { variable: Var, ports: []const Rule, set: Set }) = .empty, + vars: std.ArrayList(Var) = .empty, + /// Every port a rule set or a declaration names, once, in order. + ports: std.ArrayList([]const u8) = .empty, + + fn print(r: *const Rules, w: *std.Io.Writer) std.Io.Writer.Error!void { + for (r.items.items) |item| switch (item) { + .variable => |v| try w.print("{s}={s}\n\n", .{ v.name, v.value }), + .ports => |acts| { + for (acts) |a| try printRule(w, a); + try w.writeByte('\n'); + }, + .set => |set| { + for (set.pats) |a| try printRule(w, a); + for (set.acts) |a| try printRule(w, a); + try w.writeByte('\n'); + }, + }; + } + + fn printRule(w: *std.Io.Writer, r: Rule) std.Io.Writer.Error!void { + try w.print("{s}\t{s}\t{s}\n", .{ @tagName(r.obj), @tagName(r.verb), r.arg }); + } + + fn named(r: *const Rules, port: []const u8) bool { + for (r.ports.items) |n| if (std.mem.eql(u8, n, port)) return true; + return false; + } +}; + +/// What a message is matched with, beside the rules: the message, and what +/// the rule set's patterns found so far (match.c's Exec). +const Exec = struct { + m: *Msg, + /// The last `data matches`'s match, `$0`. + match0: []const u8 = "", + /// The span a click match took, which every other `data matches` of + /// the set must take too. + span: ?[2]usize = null, + clearclick: bool = false, + setdata: bool = false, + /// What `isfile` and `isdir` found: `$file`, `$dir`. + file: ?[]const u8 = null, + dir: ?[]const u8 = null, +}; + +/// rules.c's expand: the first word of `s` (a blank outside quotes ends +/// it), quotes rc's (`'...'`, `''` a quote), `$name` a variable's value, +/// with a message the message's (`$data`, `$0`, `$file`, ...); a `$` that +/// names nothing stays. Past `out` (Plan 9's 4 KiB ebuf), `TooLong`. +/// Answers the word and how much of `s` it took. +fn expand(r: *const Rules, e: ?*const Exec, s: []const u8, out: *[4096]u8) error{TooLong}!struct { []const u8, usize } { + var w = std.Io.Writer.fixed(out); + var i: usize = 0; + var quoting = false; + while (i < s.len and (quoting or (s[i] != ' ' and s[i] != '\t'))) { + if (s[i] == '\'') { + i += 1; + if (!quoting) { + quoting = true; + } else if (i < s.len and s[i] == '\'') { + w.writeByte('\'') catch return error.TooLong; + i += 1; + } else quoting = false; + continue; + } + if (quoting or s[i] != '$') { + w.writeByte(s[i]) catch return error.TooLong; + i += 1; + continue; + } + i += 1; + const took = dollar(r, e, s[i..], &w) catch return error.TooLong; + if (took) |n| i += n else w.writeByte('$') catch return error.TooLong; + } + return .{ w.buffered(), i }; +} + +/// rules.c's dollar: the value `$` then `s` names, written to `w`; how +/// much of `s` its name took, or null when it names nothing. +fn dollar(r: *const Rules, e: ?*const Exec, s: []const u8, w: *std.Io.Writer) std.Io.Writer.Error!?usize { + if (e) |x| if (s.len > 0 and std.ascii.isDigit(s[0])) { + // `$1`-`$9` were refused at parse (no submatches): `$0` alone. + if (s[0] == '0') try w.writeAll(x.match0); + return 1; + }; + var n: usize = 0; + while (n < s.len and std.ascii.isAlphanumeric(s[n])) n += 1; + const name = s[0..n]; + if (e) |x| { + const m = x.m; + const field: ?[]const u8 = if (std.mem.eql(u8, name, "src")) m.src else if (std.mem.eql(u8, name, "dst")) m.dst else if (std.mem.eql(u8, name, "attr")) m.attr else if (std.mem.eql(u8, name, "data")) m.data else if (std.mem.eql(u8, name, "type")) m.type else if (std.mem.eql(u8, name, "wdir")) m.wdir else null; + if (field) |f| { + try w.writeAll(f); + return n; + } + // rules.c's filename: what isfile or isdir found, else the data + // made absolute in wdir. + const found = if (std.mem.eql(u8, name, "file")) x.file else if (std.mem.eql(u8, name, "dir")) x.dir else null; + if (found != null or std.mem.eql(u8, name, "file") or std.mem.eql(u8, name, "dir")) { + if (found) |f| { + try w.writeAll(f); + } else if ((m.data.len > 0 and m.data[0] == '/') or m.wdir.len == 0) { + try w.writeAll(m.data); + } else try w.print("{s}/{s}", .{ m.wdir, m.data }); + return n; + } + } + for (r.vars.items) |v| if (std.mem.eql(u8, v.name, name)) { + try w.writeAll(v.qvalue); + return n; + }; + return null; +} + +/// Reads rule text into `rules` (rules.c's readrules): line by line, a +/// blank line (or a comment, or a variable) ending a rule set, an include +/// read in its place. A refusal is `why`: `<file>:<line>: <rules.c's words>`. +const Parser = struct { + arena: std.mem.Allocator, + /// Where an include is read from: none, only `basic`. + p: ?*Pardes, + /// Rules written, not the ones in force read again: their patterns + /// compiled and their bounds counted. + check: bool, + rules: Rules = .{}, + pats: std.ArrayList(Rule) = .empty, + acts: std.ArrayList(Rule) = .empty, + port: ?[]const u8 = null, + commands: usize = 0, + file: []const u8 = "", + line: usize = 0, + depth: usize = 0, + why: []const u8 = "", + + const Error = error{ Parse, OutOfMemory }; + + fn fail(ps: *Parser, comptime fmt: []const u8, args: anytype) Error { + ps.why = try std.fmt.allocPrint(ps.arena, "{s}:{d}: " ++ fmt, .{ ps.file, ps.line } ++ args); + return error.Parse; + } + + fn feed(ps: *Parser, name: []const u8, text: []const u8) Error!void { + ps.file = name; + ps.line = 0; + const whole = if (std.mem.endsWith(u8, text, "\n")) text[0 .. text.len - 1] else text; + var lines = std.mem.splitScalar(u8, whole, '\n'); + while (lines.next()) |line| { + ps.line += 1; + try ps.take(line); + } + } + + /// The text read, ended as its end ends a rule set. + fn finish(ps: *Parser) Error!void { + try ps.endSet(); + } + + /// rules.c's readrule, for one line. + fn take(ps: *Parser, raw: []const u8) Error!void { + const l = std.mem.trimStart(u8, raw, " \t"); + if (l.len == 0 or l[0] == '#') return ps.endSet(); + if (std.mem.startsWith(u8, l, "include")) return ps.include(l); + if (try ps.assignment(l)) return; + const obj_end = std.mem.indexOfAny(u8, l, " \t") orelse return ps.fail("malformed rule", .{}); + const obj_word = l[0..obj_end]; + const obj = std.meta.stringToEnum(Obj, obj_word) orelse + if (std.mem.eql(u8, obj_word, "kind")) .type else return ps.fail("unknown object {s}", .{obj_word}); + const rest = std.mem.trimStart(u8, l[obj_end..], " \t"); + const verb_end = std.mem.indexOfAny(u8, rest, " \t") orelse return ps.fail("malformed rule", .{}); + const verb = std.meta.stringToEnum(Verb, rest[0..verb_end]) orelse return ps.fail("unknown verb {s}", .{rest[0..verb_end]}); + const arg = std.mem.trimStart(u8, rest[verb_end..], " \t"); + if (arg.len == 0) return ps.fail("malformed rule", .{}); + // parserule. + const action = verb == .client or verb == .start or verb == .to; + if ((obj == .plumb) != action or (obj != .attr and (verb == .add or verb == .delete))) + return ps.fail("{s} not valid verb for object {s}", .{ @tagName(verb), @tagName(obj) }); + // pardes's regular expressions have no capture groups (regexp.zig, + // mvzr), and the user chose never to have an engine of pardes's + // own: a submatch is refused, never silently empty. + var quoting = false; + for (arg, 0..) |c, i| { + if (c == '\'') quoting = !quoting; + if (!quoting and c == '$' and i + 1 < arg.len and arg[i + 1] >= '1' and arg[i + 1] <= '9') + return ps.fail("${c}: no submatches (the regular expressions have no capture groups): use $0 and several matches", .{arg[i + 1]}); + } + const r: Rule = .{ .obj = obj, .verb = verb, .arg = arg, .qarg = try ps.word(arg) }; + if (verb == .matches and ps.check) { + var pat: [4096 + 4]u8 = undefined; + _ = Regex.compile(anchored(r.qarg, &pat)) catch |err| return ps.fail("{s}", .{Regex.message(err)}); + } + if (obj != .plumb) return ps.pats.append(ps.arena, r); + try ps.acts.append(ps.arena, r); + if (verb != .to) { + ps.commands += 1; + return; + } + // readruleset's: a declaration names many, a rule set one. + if (ps.pats.items.len > 0 and ps.port != null) return ps.fail("too many ports", .{}); + for ([_][]const u8{ ".", "..", "send", "rules" }) |bad| if (std.mem.eql(u8, r.qarg, bad)) + return ps.fail("illegal port name {s}", .{r.qarg}); + ps.port = r.qarg; + } + + /// readruleset's end: a rule set kept, ports declared, or a refusal. + fn endSet(ps: *Parser) Error!void { + const pats = ps.pats.items; + const acts = ps.acts.items; + const port = ps.port; + const commands = ps.commands; + ps.pats = .empty; + ps.acts = .empty; + ps.port = null; + ps.commands = 0; + if (commands > 1) return ps.fail("ruleset has more than one client or start action", .{}); + if (pats.len == 0 and acts.len == 0) return; + if (pats.len > 0 and acts.len > 0) { + try ps.rules.items.append(ps.arena, .{ .set = .{ .pats = pats, .acts = acts, .port = port } }); + } else { + if (acts.len == 0 or port == null) return ps.fail("ruleset must have patterns and actions", .{}); + for (acts) |a| if (a.verb != .to) return ps.fail("ruleset must have actions", .{}); + try ps.rules.items.append(ps.arena, .{ .ports = acts }); + } + for (acts) |a| if (a.verb == .to and !ps.rules.named(a.qarg)) { + if (ps.check and ps.rules.ports.items.len >= max_ports) return ps.fail("too many ports; max {d}", .{max_ports}); + try ps.rules.ports.append(ps.arena, a.qarg); + }; + } + + /// `s`'s first word expanded with the variables: a slice of `s` when + /// that changes nothing (no quote, no `$`), as in most rules. What a + /// parse keeps slices the text it read, which outlives it (the rules in + /// force, a write's, an include read into the arena, src/plumbing). + fn word(ps: *Parser, s: []const u8) Error![]const u8 { + var buf: [4096]u8 = undefined; + const w, const took = expand(&ps.rules, null, s, &buf) catch return ps.fail("string-too-long", .{}); + return if (std.mem.eql(u8, w, s[0..took])) s[0..took] else ps.arena.dupe(u8, w); + } + + /// rules.c's assignment: `name=value`, its value expanded once, now; + /// it ends a rule set as a blank line does. + fn assignment(ps: *Parser, l: []const u8) Error!bool { + if (!std.ascii.isAlphabetic(l[0])) return false; + var i: usize = 0; + while (i < l.len and std.ascii.isAlphanumeric(l[i])) i += 1; + const name = l[0..i]; + while (i < l.len and (l[i] == ' ' or l[i] == '\t')) i += 1; + if (i >= l.len or l[i] != '=') return false; + try ps.endSet(); + const value = std.mem.trimStart(u8, l[i + 1 ..], " \t"); + const v: Var = .{ .name = name, .value = value, .qvalue = try ps.word(value) }; + try ps.rules.items.append(ps.arena, .{ .variable = v }); + for (ps.rules.vars.items) |*old| if (std.mem.eql(u8, old.name, name)) { + old.* = v; + return true; + }; + if (ps.check and ps.rules.vars.items.len >= max_vars) return ps.fail("too many variables; max {d}", .{max_vars}); + try ps.rules.vars.append(ps.arena, v); + return true; + } + + /// rules.c's include: `include basic` is src/plumbing; another name + /// the configuration directory's file, an absolute one where it says. + fn include(ps: *Parser, l: []const u8) Error!void { + var words = std.mem.tokenizeAny(u8, l, " \t"); + const first = words.next().?; + const name = words.next() orelse ""; + const third = words.next(); + if (!std.mem.eql(u8, first, "include") or name.len == 0 or name[0] == '#' or (third != null and third.?[0] != '#')) + return ps.fail("malformed include statement", .{}); + if (ps.depth >= max_depth) return ps.fail("include stack too deep; max {d}", .{max_depth}); + const text = if (std.mem.eql(u8, name, "basic")) basic else text: { + var buf: [4096]u8 = undefined; + const p = ps.p orelse return ps.fail("can't open {s} for inclusion", .{name}); + const path = if (name[0] == '/') name else std.fmt.bufPrint(&buf, "{s}/{s}", .{ p.configDir() orelse "", name }) catch name; + break :text fs.readFile(ps.arena, path) catch return ps.fail("can't open {s} for inclusion", .{path}); + }; + const file = ps.file; + const line = ps.line; + ps.depth += 1; + try ps.feed(name, text); + ps.depth -= 1; + ps.file = file; + ps.line = line; + } +}; + +/// `pat` matching only the whole of what it is run over: `^(pat)$` (its +/// own anchors dropped, mvzr taking `^` only first), in `out`. +fn anchored(pat: []const u8, out: *[4096 + 4]u8) []const u8 { + var core = pat; + if (core.len > 0 and core[0] == '^') core = core[1..]; + if (core.len > 1 and core[core.len - 1] == '$' and core[core.len - 2] != '\\') core = core[0 .. core.len - 1]; + return std.fmt.bufPrint(out, "^({s})$", .{core}) catch pat; +} + +/// The rules in force as text to read again. +fn source(p: *Pardes) []const u8 { + return if (p.plumber.written) p.plumber.text.items else "include basic\n"; +} + +/// The rules in force read, for a send. +fn rulesOf(p: *Pardes, arena: std.mem.Allocator) Parser.Error!Rules { + var ps: Parser = .{ .arena = arena, .p = p, .check = false }; + try ps.feed("plumb/rules", source(p)); + try ps.finish(); + return ps.rules; +} + +/// The rules in force as plumb/rules reads them, the defaults printed +/// (in the step's scratch) while none were written. +pub fn inForce(p: *Pardes) []const u8 { + if (p.plumber.written) return p.plumber.text.items; + const arena = p.scratch.allocator(); + const rules = rulesOf(p, arena) catch return ""; + var out: std.Io.Writer.Allocating = .init(arena); + rules.print(&out.writer) catch return ""; + return out.written(); +} + +/// No rules at all: a `>` to plumb/rules, before what it writes. +pub fn clear(p: *Pardes) void { + p.plumber.text.clearRetainingCapacity(); + p.plumber.written = true; +} + +/// `more` read after the rules in force (`keep`) or in their place, and +/// what that makes kept as the rules: plumb/rules' write and its close's +/// rest. Null when taken, else why not, in the step's scratch, the rules +/// as they were. +pub fn apply(p: *Pardes, keep: bool, more: []const u8) ?[]const u8 { + const arena = p.scratch.allocator(); + var ps: Parser = .{ .arena = arena, .p = p, .check = true }; + const said = said: { + if (keep) { + ps.feed("plumb/rules", source(p)) catch break :said ps.why; + ps.finish() catch break :said ps.why; + } + ps.feed("plumb/rules", more) catch break :said ps.why; + ps.finish() catch break :said ps.why; + var out: std.Io.Writer.Allocating = .init(arena); + ps.rules.print(&out.writer) catch break :said "out of memory"; + if (out.written().len > max_text) break :said std.fmt.comptimePrint("rules too long; max {d} KiB", .{max_text / 1024}); + p.plumber.text.clearRetainingCapacity(); + p.plumber.text.appendSlice(p.gpa, out.written()) catch break :said "out of memory"; + p.plumber.written = true; + return null; + }; + return if (said.len > 0) said else "out of memory"; +} + +// ---- matching ---- + +/// One message's match (fsys.c's write of `send` through dispose): the +/// rules, and what the disk said of a name, asked once however many rules +/// ask it. +const Send = struct { + p: *Pardes, + /// The pane a look sent from, whose jump list's directories a name is + /// looked for in too; none from plumb/send. + from: ?usize, + arena: std.mem.Allocator, + rules: Rules, + memo_name: ?[]const u8 = null, + memo_wdir: []const u8 = "", + memo: ?fs.Resolved = null, + memo_buf: [4096]u8 = undefined, + + /// Where a name is, as Look finds it: the place written after it + /// (`:12`, `:12:5`) no part of it, then the message's directory, the + /// jump list's panes' (from a look), then Incl's (look.zig's + /// resolveLookTarget). Each stat is a taped resolve site; a name is + /// asked once a send. + fn resolve(s: *Send, wdir: []const u8, raw: []const u8) ?fs.Resolved { + const name = look.parsePathLine(std.mem.trim(u8, raw, " \t\r\n")).path; + if (s.memo_name) |n| if (std.mem.eql(u8, n, name) and std.mem.eql(u8, s.memo_wdir, wdir)) return s.memo; + s.memo_name = s.arena.dupe(u8, name) catch return null; + s.memo_wdir = s.arena.dupe(u8, wdir) catch return null; + s.memo = s.search(wdir, name); + return s.memo; + } + + fn search(s: *Send, wdir: []const u8, name: []const u8) ?fs.Resolved { + if (name.len == 0) return null; + if (fs.resolve(s.p, name, wdir, &s.memo_buf)) |found| return found; + if (name[0] == '/' or fs.dotRelative(name)) return null; + if (s.from) |id| { + const v = s.p.view(); + var seen: [pardes.MAX_PANES][]const u8 = undefined; + var nseen: usize = 0; + var n: usize = 1; + cand: while (n <= v.jumps.n) : (n += 1) { + const k = v.jumps.list[v.jumps.n - n].pane; + if (k == id) continue; + const dir = Pardes.paneDir(s.p.panes[k] orelse continue); + if (std.mem.eql(u8, dir, wdir)) continue; + for (seen[0..nseen]) |d| if (std.mem.eql(u8, d, dir)) continue :cand; + seen[nseen] = dir; + nseen += 1; + if (fs.resolve(s.p, name, dir, &s.memo_buf)) |found| return found; + } + } + for (s.p.incl.items) |dir| if (fs.resolve(s.p, name, dir, &s.memo_buf)) |found| return found; + return null; + } + + /// match.c's matchruleset: the message tried against `set`, rewritten + /// in place by its patterns (a `set` stays though a later one fails). + fn matchSet(s: *Send, set: Set, e: *Exec) error{OutOfMemory}!bool { + const m = e.m; + if (m.dst.len > 0 and set.port != null and !std.mem.eql(u8, m.dst, set.port.?)) return false; + for (set.pats) |r| if (!try s.pattern(e, r)) return false; + if (set.port) |port| if (m.dst.len == 0) { + m.dst = port; + }; + // match.c's rewrite: a click match's span is the data now. + if (e.clearclick) { + m.attr = try repack(s.arena, m.attr, "click", ""); + if (e.setdata) m.data = e.match0; + } + return true; + } + + /// match.c's matchpat. + fn pattern(s: *Send, e: *Exec, r: Rule) error{OutOfMemory}!bool { + const m = e.m; + const field: ?*[]const u8 = switch (r.obj) { + .data => &m.data, + .dst => &m.dst, + .type => &m.type, + .wdir => &m.wdir, + .src => &m.src, + .arg, .attr, .plumb => null, + }; + var buf: [4096]u8 = undefined; + switch (r.verb) { + .is => return if (field) |f| std.mem.eql(u8, f.*, r.qarg) else false, + .set => { + const f = field orelse return true; + const value = (expand(&s.rules, e, r.arg, &buf) catch return false)[0]; + f.* = try s.arena.dupe(u8, value); + if (r.obj == .data) { + e.span = null; + e.setdata = false; + } + return true; + }, + .add, .delete => { + if (r.obj != .attr) return false; + const value = (expand(&s.rules, e, r.arg, &buf) catch return false)[0]; + if (r.verb == .add) { + m.attr = try repack(s.arena, m.attr, null, value); + return true; + } + // The code's, not the man page's: an absent one fails. + var vbuf: [4096]u8 = undefined; + if (attrValue(m.attr, value, &vbuf) == null) return false; + m.attr = try repack(s.arena, m.attr, value, ""); + return true; + }, + .isfile, .isdir => { + const name = switch (r.obj) { + .arg => (expand(&s.rules, e, r.arg, &buf) catch return false)[0], + .data => m.data, + .wdir => m.wdir, + else => return false, + }; + const found = s.resolve(m.wdir, name) orelse return false; + if (found.dir != (r.verb == .isdir)) return false; + const path = try s.arena.dupe(u8, found.path); + if (r.verb == .isdir) e.dir = path else e.file = path; + return true; + }, + .matches => { + const text = (field orelse return false).*; + var pat: [4096 + 4]u8 = undefined; + var click_buf: [4096]u8 = undefined; + const click = if (r.obj == .data) attrValue(m.attr, "click", &click_buf) else null; + if (click == null) { + // The whole of it, as match.c asks of regexec's answer. + var rx = Regex.compile(anchored(r.qarg, &pat)) catch return false; + const hit = (rx.find(text, 0, 0, text.len) catch return false) orelse return false; + if (hit.start != 0 or hit.end != text.len) return false; + e.match0 = text; + return true; + } + // clickmatch: the first match from each start up to the + // click that holds it, the click counted in runes. + const runes = std.fmt.parseInt(usize, std.mem.trim(u8, click.?, " \t"), 10) catch 0; + var at: usize = 0; + var n: usize = 0; + while (n < runes and at < text.len) : (n += 1) at += std.unicode.utf8ByteSequenceLength(text[at]) catch 1; + at = @min(at, text.len); + var rx = Regex.compile(r.qarg) catch return false; + var from: usize = 0; + const hit = while (from <= at) { + const h = (rx.find(text, from, at, text.len) catch return false) orelse return false; + if (h.start <= at and at <= h.end) break h; + // Every start before the next match finds this one. + from = h.start + 1; + while (from < text.len and text[from] & 0xC0 == 0x80) from += 1; + } else return false; + if (e.span) |sp| if (sp[0] != hit.start or sp[1] != hit.end) return false; + e.span = .{ hit.start, hit.end }; + e.clearclick = true; + e.setdata = true; + e.match0 = text[hit.start..hit.end]; + return true; + }, + .client, .start, .to => return false, + } + } +}; + +/// The ports pardes reads itself, as acme reads `edit`; and every loaded +/// kind by its name. +const Port = enum { + /// A file or directory opened as Look opens one, at the place written + /// after its name or given as `addr=`. + edit, + /// A URL given to the host's browser (xdg-open, open on macOS). + web, +}; + +/// A message matched against the rules and sent where they say (fsys.c's +/// write of `send`, then dispose), from pane `from` (a look's) or from +/// plumb/send (null: the pane with the keyboard runs what it starts). +/// Null when it went somewhere, else why not, in Plan 9's words, in the +/// step's scratch. +pub fn send(p: *Pardes, from: ?usize, msg: Msg) ?[]const u8 { + const arena = p.scratch.allocator(); + var s: Send = .{ .p = p, .from = from, .arena = arena, .rules = rulesOf(p, arena) catch return "plumb/rules: the rules in force do not read" }; + var m = msg; + var e: Exec = .{ .m = &m }; + const set: ?Set = for (s.rules.items.items) |item| { + const candidate = switch (item) { + .set => |x| x, + else => continue, + }; + e = .{ .m = &m }; + if (s.matchSet(candidate, &e) catch return "out of memory") break candidate; + } else null; + const id = from orelse p.view().active; + // dispose. + if (m.dst.len == 0) return startup(&s, id, set orelse return e_no_port, &e); + if (std.meta.stringToEnum(Port, m.dst)) |port| { + deliver(&s, id, port, &m); + return null; + } + if (p.kinds.byName(m.dst)) |k| { + openInKind(&s, id, k, &m); + return null; + } + // A port nobody reads: what the rule set starts, if it is one a rule + // names; one no rule names is no port, and the message goes nowhere, + // as fsys.c's dispose finds no Dirtab for it. + if (!s.rules.named(m.dst)) return null; + return startup(&s, id, set orelse return e_no_start, &e); +} + +/// match.c's startup: the rule set's `start` or `client`, its words +/// expanded, run. +fn startup(s: *Send, id: usize, set: Set, e: *const Exec) ?[]const u8 { + const act = for (set.acts) |a| { + if (a.verb == .start) break a; + if (a.verb == .client) { + if (e.m.dst.len == 0) return e_client; + break a; + } + } else return e_no_start; + // buildargv: one expansion a word. + var argv: std.ArrayList([]const u8) = .empty; + var rest = act.arg; + while (true) { + rest = std.mem.trimStart(u8, rest, " \t"); + if (rest.len == 0) break; + var buf: [4096]u8 = undefined; + const word, const took = expand(&s.rules, e, rest, &buf) catch return "string-too-long"; + argv.append(s.arena, s.arena.dupe(u8, word) catch return "out of memory") catch return "out of memory"; + rest = rest[took..]; + } + if (argv.items.len == 0) return e_empty; + run(s.p, id, e.m.wdir, argv.items); + return null; +} + +/// What a `plumb start` (and, until ports are read, a `plumb client`) runs: +/// the one place that decides it (question 4 of docs/next-plumber.md). +/// Its words as pardes's exec runs a line from the message's pane: a +/// builtin or a kind's word in place, any other line a command pane in +/// `wdir`, each word one argument to $SHELL -c however it is spelled. +/// Plan 9's runs detached, with no window. +fn run(p: *Pardes, id: usize, wdir: []const u8, argv: []const []const u8) void { + const arena = p.scratch.allocator(); + const line = std.mem.join(arena, " ", argv) catch return p.reportError(id, "plumb", error.OutOfMemory); + if (p.executeBuiltinLine(id, line)) return; + var quoted: std.Io.Writer.Allocating = .init(arena); + for (argv, 0..) |a, i| { + if (i > 0) quoted.writer.writeByte(' ') catch return; + quoted.writer.writeByte('\'') catch return; + for (a) |c| { + if (c == '\'') { + quoted.writer.writeAll("'\\''") catch return; + } else quoted.writer.writeByte(c) catch return; + } + quoted.writer.writeByte('\'') catch return; + } + p.exec_dir = if (wdir.len > 0) wdir else null; + defer p.exec_dir = null; + _ = exec.runCommand(p, id, quoted.written()); +} + +/// What pardes does with a message to a port it reads. Like a message +/// queued to a reader in Plan 9, the send is done: what goes wrong after +/// is said as a look's failure is. +fn deliver(s: *Send, id: usize, port: Port, m: *const Msg) void { + const p = s.p; + switch (port) { + .web => { + if (m.data.len > 256) return p.reportFailure(id, "plumb: web: a URL is at most 256 bytes"); + p.emit(.{ .open_link = .from(m.data) }); + }, + .edit => { + // acme's plumblook: the data a name in wdir, at its addr. + var buf: [4096]u8 = undefined; + const addr = attrValue(m.attr, "addr", &buf) orelse ""; + const relative = m.data.len > 0 and m.data[0] != '/' and m.data[0] != '~' and m.wdir.len > 0; + const text = std.fmt.allocPrint(s.arena, "{s}{s}{s}{s}{s}", .{ + if (relative) m.wdir else "", if (relative) "/" else "", m.data, + if (addr.len > 0) ":" else "", addr, + }) catch return p.reportError(id, "plumb", error.OutOfMemory); + look.lookAt(p, id, text); + }, + } +} + +/// A message to a loaded kind's port: its data, a file, opened in that kind +/// (or its pane given the keyboard, when one shows it). +fn openInKind(s: *Send, id: usize, k: u8, m: *const Msg) void { + const p = s.p; + const found = s.resolve(m.wdir, m.data) orelse return p.reportFailure(id, "plumb: no such file"); + if (found.dir) return p.reportFailure(id, "plumb: a directory is no kind's"); + const path = s.arena.dupe(u8, found.path) catch return; + for (p.panes, 0..) |slot, i| if (slot) |pn| if (pn.addon) |a| if (std.mem.eql(u8, a.path, path)) { + p.view().active = i; + return; + }; + const free = p.freeSlot() orelse return p.reportError(id, "plumb", error.NoPaneSlots); + const made = Kind.openFile(p, free, k, path) catch |err| return p.reportError(id, path, err); + if (made.addon.?.refused.len > 0) { + made.vweight = 0; + p.unplaced.set(free); + return p.reportFailure(id, s.arena.dupe(u8, made.addon.?.refused.slice()) catch "plumb: refused"); + } + layout.placeNew(p, id, id, free, .doc); +} + +/// Where rule text written in pieces is whole: through its last blank line +/// (rules.c's morerules), 0 when none is. +pub fn wholeSets(text: []const u8) usize { + return if (std.mem.lastIndexOf(u8, text, "\n\n")) |at| at + 2 else 0; +} + +// ---- tests ---- + +const testing = std.testing; + +/// The rules `text` read as written ones are, printed; or the refusal. +fn reread(arena: std.mem.Allocator, text: []const u8) ![]const u8 { + var ps: Parser = .{ .arena = arena, .p = null, .check = true }; + ps.feed("t", text) catch return ps.why; + ps.finish() catch return ps.why; + var out: std.Io.Writer.Allocating = .init(arena); + try ps.rules.print(&out.writer); + return out.written(); +} + +test "plumb: a message packs as plumb(6)'s six lines and its data, and unpacks whole, short of its data, or malformed" { + var buf: [256]u8 = undefined; + var w = std.Io.Writer.fixed(&buf); + const m: Msg = .{ .src = "pardes", .wdir = "/tmp", .type = "text", .attr = "click=3 addr=12", .data = "a b\nc" }; + try m.pack(&w); + try testing.expectEqualStrings("pardes\n\n/tmp\ntext\nclick=3 addr=12\n5\na b\nc", w.buffered()); + const back = unpack(w.buffered()).msg; + inline for (.{ "src", "dst", "wdir", "type", "attr", "data" }) |f| try testing.expectEqualStrings(@field(m, f), @field(back, f)); + try testing.expectEqual(@as(usize, 2), unpack(w.buffered()[0 .. w.buffered().len - 2]).more); + // A header line short is no message at all, as is data past its count. + try testing.expect(unpack("pardes\n\n/tmp\ntext\n") == .bad); + try testing.expect(unpack("a\nb\nc\nd\ne\n1\nxy") == .bad); + try testing.expect(unpack("a\nb\nc\nd\ne\nzz\n") == .msg); +} + +test "plumb: attributes read rc's quotes, a value that needs them is quoted again, delete takes the first, add appends" { + var arena: std.heap.ArenaAllocator = .init(testing.allocator); + defer arena.deinit(); + const a = arena.allocator(); + var buf: [4096]u8 = undefined; + const attr = "action=showdata filename='/a b/it''s' click=4"; + try testing.expectEqualStrings("/a b/it's", attrValue(attr, "filename", &buf).?); + try testing.expectEqualStrings("4", attrValue(attr, "click", &buf).?); + try testing.expect(attrValue(attr, "addr", &buf) == null); + try testing.expectEqualStrings("action=showdata filename='/a b/it''s'", try repack(a, attr, "click", "")); + try testing.expectEqualStrings("action=showdata filename='/a b/it''s' click=4 addr='x=1'", try repack(a, attr, null, "addr='x=1'")); + // One with no `=` ends them, as plumbunpackattr stops there. + try testing.expect(attrValue("a=1 bogus b=2", "b", &buf) == null); +} + +test "plumb: every parse error in rules.c's words, with its line" { + var arena: std.heap.ArenaAllocator = .init(testing.allocator); + defer arena.deinit(); + const a = arena.allocator(); + for ([_][2][]const u8{ + .{ "type is text\nbogus\n", "t:2: malformed rule" }, + .{ "type is\n", "t:1: malformed rule" }, + .{ "type \n", "t:1: malformed rule" }, + .{ "colour is red\n", "t:1: unknown object colour" }, + .{ "type equals text\n", "t:1: unknown verb equals" }, + .{ "data to edit\n", "t:1: to not valid verb for object data" }, + .{ "type start x\n", "t:1: start not valid verb for object type" }, + .{ "data add x=1\n", "t:1: add not valid verb for object data" }, + .{ "plumb is edit\n", "t:1: is not valid verb for object plumb" }, + .{ "type is text\nplumb to a\nplumb to b\n", "t:3: too many ports" }, + .{ "type is text\nplumb to send\n", "t:2: illegal port name send" }, + .{ "plumb to ..\n", "t:1: illegal port name .." }, + .{ "plumb to rules\n", "t:1: illegal port name rules" }, + .{ "type is text\nplumb start a\nplumb client b\n", "t:3: ruleset has more than one client or start action" }, + .{ "type is text\n\n", "t:2: ruleset must have patterns and actions" }, + .{ "plumb start x\n", "t:1: ruleset must have patterns and actions" }, + .{ "plumb to a\nplumb start x\n\n", "t:3: ruleset must have actions" }, + .{ "include\n", "t:1: malformed include statement" }, + .{ "include a b\n", "t:1: malformed include statement" }, + .{ "includes x\n", "t:1: malformed include statement" }, + .{ "include nowhere\n", "t:1: can't open nowhere for inclusion" }, + .{ "data matches '[a-'\nplumb to edit\n", "t:1: bad regular expression" }, + .{ "data matches '([a-z]+)\\.c'\narg isfile $1\nplumb to edit\n", "t:2: $1: no submatches (the regular expressions have no capture groups): use $0 and several matches" }, + }) |c| try testing.expectEqualStrings(c[1], try reread(a, c[0])); + // Inside quotes a `$1` is the text `$1`. + try testing.expectEqualStrings("data\tis\t'$1'\nplumb\tto\tedit\n\n", try reread(a, "data is '$1'\nplumb to edit\n")); + // The most variables, and ports, are bounded. + var many: std.Io.Writer.Allocating = .init(a); + for (0..max_vars + 1) |i| try many.writer.print("v{d}=x\n", .{i}); + try testing.expectEqualStrings("t:33: too many variables; max 32", try reread(a, many.written())); + many.clearRetainingCapacity(); + for (0..max_ports + 1) |i| try many.writer.print("plumb to p{d}\n", .{i}); + try testing.expectEqualStrings("t:33: too many ports; max 32", try reread(a, many.written())); +} + +test "plumb: rules print as rule sets of tab-separated lines, variables where they were set, and read back the same" { + var arena: std.heap.ArenaAllocator = .init(testing.allocator); + defer arena.deinit(); + const a = arena.allocator(); + const text = + \\# a comment ends nothing it is not in + \\x = '[a-z]+' + \\plumb to seemail + \\plumb to showmail + \\ + \\kind is text + \\data matches $x + \\plumb to edit + \\x=other + \\type is text + \\data set $x + \\plumb start Msg $data + ; + const printed = try reread(a, text); + try testing.expectEqualStrings("x='[a-z]+'\n\nplumb\tto\tseemail\nplumb\tto\tshowmail\n\ntype\tis\ttext\ndata\tmatches\t$x\nplumb\tto\tedit\n\nx=other\n\ntype\tis\ttext\ndata\tset\t$x\nplumb\tstart\tMsg $data\n\n", printed); + try testing.expectEqualStrings(printed, try reread(a, printed)); + // The default rules read, and print as they read. + const defaults = try reread(a, "include basic\n"); + try testing.expect(std.mem.indexOf(u8, defaults, "data\tmatches\t'https?://.*'\nplumb\tto\tweb\n") != null); + try testing.expectEqualStrings(defaults, try reread(a, defaults)); +} + +test "plumb: rc's quoting, variables and the message's in an expansion, the first word alone" { + var arena: std.heap.ArenaAllocator = .init(testing.allocator); + defer arena.deinit(); + var ps: Parser = .{ .arena = arena.allocator(), .p = null, .check = true }; + try ps.feed("t", "v='a b'\nw=x$v'y'\n"); + var m: Msg = .{ .src = "pardes", .wdir = "/w", .data = "f.c" }; + var e: Exec = .{ .m = &m, .match0 = "zero" }; + var buf: [4096]u8 = undefined; + for ([_][2][]const u8{ + .{ "'it''s' rest", "it's" }, + .{ "pre$v.post", "prea b.post" }, + .{ "$w", "xa by" }, + .{ "'$v'", "$v" }, + .{ "$nothere$", "$nothere$" }, + .{ "$src:$wdir:$data:$0:$file", "pardes:/w:f.c:zero:/w/f.c" }, + }) |c| try testing.expectEqualStrings(c[1], (try expand(&ps.rules, &e, c[0], &buf))[0]); + e.file = "/real/f.c"; + try testing.expectEqualStrings("/real/f.c", (try expand(&ps.rules, &e, "$file", &buf))[0]); + // Past 4 KiB: too long, never cut. + const long = try arena.allocator().alloc(u8, 5000); + @memset(long, 'a'); + m.data = long; + try testing.expectError(error.TooLong, expand(&ps.rules, &e, "$data", &buf)); +} + +/// A send against `text`'s rules on a session of its own: the message +/// after its match, the rule set that took it (its index among the sets, +/// null for none) and what dispose said. +const Tried = struct { m: Msg, set: ?usize }; + +fn tryRules(p: *Pardes, text: []const u8, msg: Msg) !Tried { + try testing.expect(apply(p, false, text) == null); + const arena = p.scratch.allocator(); + var s: Send = .{ .p = p, .from = null, .arena = arena, .rules = try rulesOf(p, arena) }; + var m = msg; + var e: Exec = .{ .m = &m }; + var n: usize = 0; + for (s.rules.items.items) |item| { + const set = switch (item) { + .set => |x| x, + else => continue, + }; + e = .{ .m = &m }; + if (try s.matchSet(set, &e)) return .{ .m = m, .set = n }; + n += 1; + } + return .{ .m = m, .set = null }; +} + +test "plumb: matches takes the whole text, or with a click the run around it: Pike's horse.gift is no gif" { + const p = try Pardes.init(testing.allocator, .{ .cols = 80, .rows = 24 }); + defer p.deinit(); + const rules = + \\type is text + \\data matches '[a-zA-Z0-9_\-./]+' + \\data matches '([a-zA-Z0-9_\-./]+)\.(gif|jpe?g)' + \\plumb to image + \\ + ; + try testing.expectEqual(@as(?usize, 0), (try tryRules(p, rules, .{ .type = "text", .data = "horse.gif" })).set); + try testing.expect((try tryRules(p, rules, .{ .type = "text", .data = "horse.gift" })).set == null); + try testing.expect((try tryRules(p, rules, .{ .type = "text", .data = "see horse.gif" })).set == null); + // A click inside the name: the run of file characters around it, the + // data cut to it and the click gone. + const hit = try tryRules(p, rules, .{ .type = "text", .attr = "click=7", .data = "see horse.gif now" }); + try testing.expectEqual(@as(?usize, 0), hit.set); + try testing.expectEqualStrings("horse.gif", hit.m.data); + try testing.expectEqualStrings("", hit.m.attr); + try testing.expectEqualStrings("image", hit.m.dst); + try testing.expect((try tryRules(p, rules, .{ .type = "text", .attr = "click=7", .data = "see horse.gift now" })).set == null); + // The click counts runes: two bytes of é before it. + try testing.expectEqualStrings("horse.gif", (try tryRules(p, rules, .{ .type = "text", .attr = "click=3", .data = "é horse.gif" })).m.data); + // Each object a rule matches whole. + const src = "src\tmatches\t'par.*'\ntype\tmatches\t'te(x|y)t'\nplumb\tto\tedit\n"; + try testing.expectEqual(@as(?usize, 0), (try tryRules(p, src, .{ .src = "pardes", .type = "text" })).set); + try testing.expect((try tryRules(p, src, .{ .src = "pardes", .type = "texts" })).set == null); + // The default rules' URL: http and https, whole. + try testing.expectEqual(@as(?usize, 0), (try tryRules(p, basic, .{ .type = "text", .data = "https://ziglang.org/x?y" })).set); + try testing.expect((try tryRules(p, basic, .{ .type = "text", .data = "see https://ziglang.org" })).set == null); +} + +test "plumb: a set rewrites the message for good, attr delete of one absent fails, a message to a port tries only rule sets to it" { + const p = try Pardes.init(testing.allocator, .{ .cols = 80, .rows = 24 }); + defer p.deinit(); + const rules = + \\data set changed + \\type is never + \\plumb to a + \\ + \\attr delete action + \\attr add 'action=showdiff filename=x' + \\plumb to b + \\ + \\data is changed + \\plumb to c + \\ + ; + // The first set fails after its `set`: the data stays changed; the + // second fails at its delete; the third sees the changed data. + const t = try tryRules(p, rules, .{ .type = "text", .data = "orig" }); + try testing.expectEqual(@as(?usize, 2), t.set); + try testing.expectEqualStrings("changed", t.m.data); + try testing.expectEqualStrings("c", t.m.dst); + const d = try tryRules(p, rules, .{ .type = "text", .attr = "action=showdata", .data = "x" }); + try testing.expectEqual(@as(?usize, 1), d.set); + try testing.expectEqualStrings("action=showdiff filename=x", d.m.attr); + // dst names a port: rule sets to another are passed over. + const to_a = try tryRules(p, "data\tis\tx\nplumb\tto\ta\n\ndata\tis\tx\nplumb\tto\tb\n\n", .{ .dst = "b", .data = "x" }); + try testing.expectEqual(@as(?usize, 1), to_a.set); +} + +test "plumb: isfile and isdir find a name as Look does, past its place, in the message's directory then Incl's; $file is where" { + if (!fs.platform_has_fs) return; + var tmp = testing.tmpDir(.{}); + defer tmp.cleanup(); + try tmp.dir.writeFile(testing.io, .{ .sub_path = "it.zig", .data = "x\n" }); + try tmp.dir.createDir(testing.io, "sub", .default_dir); + var dir_buf: [4096]u8 = undefined; + const root = dir_buf[0..try tmp.dir.realPath(testing.io, &dir_buf)]; + const p = try Pardes.init(testing.allocator, .{ .cols = 80, .rows = 24 }); + defer p.deinit(); + const rules = "arg\tisfile\t$data\ndata\tset\t$file\nplumb\tto\tedit\n\narg\tisdir\t$data\ndata\tset\t$dir\nplumb\tto\tedit\n\n"; + const file = try tryRules(p, rules, .{ .wdir = root, .data = "it.zig:3:2" }); + try testing.expectEqual(@as(?usize, 0), file.set); + try testing.expect(std.mem.endsWith(u8, file.m.data, "/it.zig")); + const dir = try tryRules(p, rules, .{ .wdir = root, .data = "sub" }); + try testing.expectEqual(@as(?usize, 1), dir.set); + try testing.expect(std.mem.endsWith(u8, dir.m.data, "/sub")); + try testing.expect((try tryRules(p, rules, .{ .wdir = root, .data = "nothere.zig" })).set == null); + // Not in wdir: Incl's directories, as acme's includename. + try p.incl.append(p.gpa, try p.gpa.dupe(u8, root)); + try testing.expectEqual(@as(?usize, 0), (try tryRules(p, rules, .{ .wdir = "/", .data = "it.zig" })).set); + try testing.expect((try tryRules(p, rules, .{ .wdir = "/", .data = "./it.zig" })).set == null); +} + +test "plumb: dispose: no rule and no dst, a start, the web port, a port no rule names" { + const p = try Pardes.init(testing.allocator, .{ .cols = 80, .rows = 24 }); + defer p.deinit(); + _ = try p.setTestFile("x\n"); + try testing.expect(apply(p, false, "include basic\n\ntype\tis\tmsg\nplumb\tstart\tMsg $data\n\ntype\tis\tnone\nplumb\tto\tnobody\n\ntype\tis\tclient\nplumb\tclient\tMsg x\n\n") == null); + try testing.expectEqualStrings(e_no_port, send(p, null, .{ .type = "text", .data = "no such thing" }).?); + // A start of a builtin runs it in place. + try testing.expect(send(p, null, .{ .type = "msg", .data = "hello" }) == null); + const pane = p.panes[p.view().active].?; + try testing.expectEqualStrings("hello", pane.msg[0..pane.msg_len]); + try testing.expectEqualStrings(e_client, send(p, null, .{ .type = "client" }).?); + try testing.expectEqualStrings(e_no_start, send(p, null, .{ .type = "none" }).?); + try testing.expectEqualStrings(e_no_start, send(p, null, .{ .dst = "nobody", .type = "x" }).?); + // A port no rule names goes nowhere, said nothing (fsys.c's dispose). + try testing.expect(send(p, null, .{ .dst = "elsewhere", .data = "x" }) == null); + // A URL to the web port: the host's opener. + while (p.nextEffect()) |_| {} + try testing.expect(send(p, null, .{ .type = "text", .data = "https://ziglang.org" }) == null); + var opened = false; + while (p.nextEffect()) |e| switch (e) { + .open_link => |u| { + try testing.expectEqualStrings("https://ziglang.org", u.slice()); + opened = true; + }, + else => {}, + }; + try testing.expect(opened); +} + +test "plumb: rules written replace or add, a refusal keeps them as they were, and a dump keeps them" { + const p = try Pardes.init(testing.allocator, .{ .cols = 80, .rows = 24 }); + defer p.deinit(); + // Every session starts with the defaults. + try testing.expect(std.mem.indexOf(u8, inForce(p), "plumb\tto\tweb") != null); + try testing.expect(!p.plumber.written); + try testing.expect(apply(p, false, "type is a\nplumb to edit\n") == null); + try testing.expect(apply(p, true, "type is b\nplumb to edit\n") == null); + try testing.expectEqualStrings("type\tis\ta\nplumb\tto\tedit\n\ntype\tis\tb\nplumb\tto\tedit\n\n", p.plumber.text.items); + try testing.expectEqualStrings("plumb/rules:1: unknown verb isnt", apply(p, true, "type isnt c\n").?); + try testing.expectEqualStrings("type\tis\ta\nplumb\tto\tedit\n\ntype\tis\tb\nplumb\tto\tedit\n\n", p.plumber.text.items); + try pardes.dump.dumpState(p); + const restored = try pardes.dump.restore(p, p.dump_out.?, "/tmp/pardes.dump.zon"); + defer restored.deinit(); + try testing.expectEqualStrings(p.plumber.text.items, restored.plumber.text.items); + try testing.expect(restored.plumber.written); +} + +test "plumb: plan9port's basic parses, its ¡- classes UTF-8's bytes and its submatch rule sets left out, and its rules match" { + var arena: std.heap.ArenaAllocator = .init(testing.allocator); + defer arena.deinit(); + // plan9port's plumb/basic and fileaddr; `¡-` is `\x80-\xf4`, the bytes + // of UTF-8's multi-byte runes (a run of them is whole runes: UTF-8 is + // self-synchronising), and the rule sets that need `$1`-`$5` are gone. + const p9p = + \\addrelem='((#?[0-9]+)|(/[A-Za-z0-9_\^]+/?)|[.$])' + \\addr=:($addrelem([,;+\-]$addrelem)*) + \\ + \\plumb to seemail + \\plumb to showmail + \\ + \\type is text + \\data matches '(https?|ftp|file|gopher|mailto|news|nntp|telnet|wais|prospero)://[a-zA-Z0-9_@\-]+([.:][a-zA-Z0-9_@\-]+)*/?[a-zA-Z0-9_?,%#~&/\-+=]+([:.][a-zA-Z0-9_?,%#~&/\-+=]+)*' + \\plumb to web + \\plumb start web $0 + \\ + \\type is text + \\data matches '[a-zA-Z\x80-\xf40-9_\-./]+' + \\data matches '([a-zA-Z\x80-\xf40-9_\-./]+)\.(doc|rtf)' + \\arg isfile $0 + \\plumb to msword + \\plumb start wdoc2txt $file + \\ + \\type is text + \\dst is msword + \\plumb to msword + \\plumb start wdoc2txt $file + \\ + \\type is text + \\data matches '[a-zA-Z\x80-\xf40-9_\-./@]+' + \\data matches '([a-zA-Z\x80-\xf40-9_\-./@]+)\.(jpe?g|JPE?G|gif|GIF|tiff?|TIFF?|ppm|bit|png|PNG)' + \\arg isfile $0 + \\plumb to image + \\plumb start 9 page $file + \\ + \\type is text + \\data matches '[a-zA-Z\x80-\xf40-9_\-./@]+' + \\data matches '([a-zA-Z\x80-\xf40-9_\-./@]+)\.(ps|PS|eps|EPS|pdf|PDF|dvi|DVI)' + \\arg isfile $0 + \\plumb to postscript + \\plumb start 9 page $file + \\ + \\data matches '[a-zA-Z\x80-\xf40-9_\-./@]+' + \\data matches '([a-zA-Z\x80-\xf40-9_\-./@]+)\.([Ss][XxTt][CcDdIiGgMmWw]|[Dd][Oo][Cc]|[Xx][Ll][Ss]|[Pp][Pp][Tt])' + \\arg isfile $0 + \\plumb to openoffice + \\plumb start openoffice $file + \\ + \\type is text + \\data matches '[a-zA-Z\x80-\xf40-9_\-./]+' + \\data matches '/mail/fs/[a-zA-Z\x80-\xf40-9/]+/[0-9]+' + \\plumb to showmail + \\plumb start window -r 4 120 750 600 upas/nedmail -s $0 + \\ + \\type is text + \\data matches '[a-zA-Z0-9_+.\-]+@[a-zA-Z0-9_+.\-]*' + \\plumb to sendmail + \\plumb start wmail $0 + \\ + \\dst is image + \\arg isfile $data + \\plumb to image + \\plumb start 9 page $data + \\ + \\dst is postscript + \\arg isfile $data + \\plumb start 9 page $data + \\ + ; + const printed = try reread(arena.allocator(), p9p); + try testing.expect(std.mem.startsWith(u8, printed, "addrelem='((#?[0-9]+)|(/[A-Za-z0-9_\\^]+/?)|[.$])'\n\naddr=:($addrelem([,;+\\-]$addrelem)*)\n\nplumb\tto\tseemail\n")); + try testing.expectEqualStrings(printed, try reread(arena.allocator(), printed)); + const p = try Pardes.init(testing.allocator, .{ .cols = 80, .rows = 24 }); + defer p.deinit(); + // Clicked in a line: the URL's run, the mail address's, and the + // faces' path, whole runes of a name that is not ASCII. + const url = try tryRules(p, p9p, .{ .type = "text", .attr = "click=10", .data = "see https://9fans.github.io/plan9port/ now" }); + try testing.expectEqual(@as(?usize, 0), url.set); + try testing.expectEqualStrings("https://9fans.github.io/plan9port/", url.m.data); + try testing.expectEqualStrings("web", url.m.dst); + const mail = try tryRules(p, p9p, .{ .type = "text", .attr = "click=12", .data = "write to [email protected]." }); + try testing.expectEqualStrings("sendmail", mail.m.dst); + try testing.expectEqualStrings("[email protected].", mail.m.data); + const face = try tryRules(p, p9p, .{ .type = "text", .attr = "click=3", .data = "x /mail/fs/café/12 y" }); + try testing.expectEqualStrings("showmail", face.m.dst); + try testing.expectEqualStrings("/mail/fs/café/12", face.m.data); + // An image named in UTF-8, and Pike's horse.gift, by the image set's + // two patterns alone. + const pics = "data matches '[a-zA-Z\\x80-\\xf40-9_\\-./@]+'\ndata matches '([a-zA-Z\\x80-\\xf40-9_\\-./@]+)\\.(jpe?g|gif|png)'\nplumb to image\n"; + try testing.expectEqualStrings("Ångström.png", (try tryRules(p, pics, .{ .attr = "click=2", .data = "Ångström.png" })).m.data); + try testing.expect((try tryRules(p, pics, .{ .attr = "click=2", .data = "horse.gift" })).set == null); + // The rune range as written is refused, in regexp.zig's words. + try testing.expectEqualStrings("t:1: bad regular expression: a range of runes wider than 256 in [...] is not supported", try reread(arena.allocator(), "data matches '[a-zA-Z¡-0-9_\\-./]+'\nplumb to edit\n")); +} |
