Merge pull request #10456 from Aransentin/master

Support parsing tz timezone data

5 files changed, 253 insertions, 0 deletions

diff --git a/lib/std/std.zig b/lib/std/std.zig
index f94185f2fb..5ae09a7f5b 100644
--- a/lib/std/std.zig
+++ b/lib/std/std.zig
@@ -39,6 +39,7 @@ pub const StringArrayHashMapUnmanaged = array_hash_map.StringArrayHashMapUnmanag
 pub const TailQueue = @import("linked_list.zig").TailQueue;
 pub const Target = @import("target.zig").Target;
 pub const Thread = @import("Thread.zig");
+pub const Tz = @import("tz.zig").Tz;
 
 pub const array_hash_map = @import("array_hash_map.zig");
 pub const atomic = @import("atomic.zig");
diff --git a/lib/std/tz.zig b/lib/std/tz.zig
new file mode 100644
index 0000000000..d505a790b7
--- /dev/null
+++ b/lib/std/tz.zig
@@ -0,0 +1,252 @@
+const std = @import("std.zig");
+const builtin = @import("builtin");
+
+pub const Transition = struct {
+    ts: i64,
+    timetype: *Timetype,
+};
+
+pub const Timetype = struct {
+    offset: i32,
+    flags: u8,
+    name_data: [6:0]u8,
+
+    pub fn name(self: Timetype) [:0]const u8 {
+        return std.mem.sliceTo(self.name_data[0..], 0);
+    }
+
+    pub fn isDst(self: Timetype) bool {
+        return (self.flags & 0x01) > 0;
+    }
+
+    pub fn standardTimeIndicator(self: Timetype) bool {
+        return (self.flags & 0x02) > 0;
+    }
+
+    pub fn utIndicator(self: Timetype) bool {
+        return (self.flags & 0x04) > 0;
+    }
+};
+
+pub const Leapsecond = struct {
+    occurrence: i48,
+    correction: i16,
+};
+
+pub const Tz = struct {
+    allocator: std.mem.Allocator,
+    transitions: []const Transition,
+    timetypes: []const Timetype,
+    leapseconds: []const Leapsecond,
+    footer: ?[]const u8,
+
+    const Header = extern struct {
+        magic: [4]u8,
+        version: u8,
+        reserved: [15]u8,
+        counts: extern struct {
+            isutcnt: u32,
+            isstdcnt: u32,
+            leapcnt: u32,
+            timecnt: u32,
+            typecnt: u32,
+            charcnt: u32,
+        },
+    };
+
+    pub fn parse(allocator: std.mem.Allocator, reader: anytype) !Tz {
+        var legacy_header = try reader.readStruct(Header);
+        if (!std.mem.eql(u8, &legacy_header.magic, "TZif")) return error.BadHeader;
+        if (legacy_header.version != 0 and legacy_header.version != '2' and legacy_header.version != '3') return error.BadVersion;
+
+        if (builtin.target.cpu.arch.endian() != std.builtin.Endian.Big) {
+            std.mem.bswapAllFields(@TypeOf(legacy_header.counts), &legacy_header.counts);
+        }
+
+        if (legacy_header.version == 0) {
+            return parseBlock(allocator, reader, legacy_header, true);
+        } else {
+            // If the format is modern, just skip over the legacy data
+            const skipv = legacy_header.counts.timecnt * 5 + legacy_header.counts.typecnt * 6 + legacy_header.counts.charcnt + legacy_header.counts.leapcnt * 8 + legacy_header.counts.isstdcnt + legacy_header.counts.isutcnt;
+            try reader.skipBytes(skipv, .{});
+
+            var header = try reader.readStruct(Header);
+            if (!std.mem.eql(u8, &header.magic, "TZif")) return error.BadHeader;
+            if (header.version != '2' and header.version != '3') return error.BadVersion;
+            if (builtin.target.cpu.arch.endian() != std.builtin.Endian.Big) {
+                std.mem.bswapAllFields(@TypeOf(header.counts), &header.counts);
+            }
+
+            return parseBlock(allocator, reader, header, false);
+        }
+    }
+
+    fn parseBlock(allocator: std.mem.Allocator, reader: anytype, header: Header, legacy: bool) !Tz {
+        if (header.counts.isstdcnt != 0 and header.counts.isstdcnt != header.counts.typecnt) return error.Malformed; // rfc8536: isstdcnt [...] MUST either be zero or equal to "typecnt"
+        if (header.counts.isutcnt != 0 and header.counts.isutcnt != header.counts.typecnt) return error.Malformed; // rfc8536: isutcnt [...] MUST either be zero or equal to "typecnt"
+        if (header.counts.typecnt == 0) return error.Malformed; // rfc8536: typecnt [...] MUST NOT be zero
+        if (header.counts.charcnt == 0) return error.Malformed; // rfc8536: charcnt [...] MUST NOT be zero
+        if (header.counts.charcnt > 256 + 6) return error.Malformed; // Not explicitly banned by rfc8536 but nonsensical
+
+        var leapseconds = try allocator.alloc(Leapsecond, header.counts.leapcnt);
+        errdefer allocator.free(leapseconds);
+        var transitions = try allocator.alloc(Transition, header.counts.timecnt);
+        errdefer allocator.free(transitions);
+        var timetypes = try allocator.alloc(Timetype, header.counts.typecnt);
+        errdefer allocator.free(timetypes);
+
+        // Parse transition types
+        var i: usize = 0;
+        while (i < header.counts.timecnt) : (i += 1) {
+            transitions[i].ts = if (legacy) try reader.readIntBig(i32) else try reader.readIntBig(i64);
+        }
+
+        i = 0;
+        while (i < header.counts.timecnt) : (i += 1) {
+            const tt = try reader.readByte();
+            if (tt >= timetypes.len) return error.Malformed; // rfc8536: Each type index MUST be in the range [0, "typecnt" - 1]
+            transitions[i].timetype = &timetypes[tt];
+        }
+
+        // Parse time types
+        i = 0;
+        while (i < header.counts.typecnt) : (i += 1) {
+            const offset = try reader.readIntBig(i32);
+            if (offset < -2147483648) return error.Malformed; // rfc8536: utoff [...] MUST NOT be -2**31
+            const dst = try reader.readByte();
+            if (dst != 0 and dst != 1) return error.Malformed; // rfc8536: (is)dst [...] The value MUST be 0 or 1.
+            const idx = try reader.readByte();
+            if (idx > header.counts.charcnt - 1) return error.Malformed; // rfc8536: (desig)idx [...] Each index MUST be in the range [0, "charcnt" - 1]
+            timetypes[i] = .{
+                .offset = offset,
+                .flags = dst,
+                .name_data = undefined,
+            };
+
+            // Temporarily cache idx in name_data to be processed after we've read the designator names below
+            timetypes[i].name_data[0] = idx;
+        }
+
+        var designators_data: [256 + 6]u8 = undefined;
+        try reader.readNoEof(designators_data[0..header.counts.charcnt]);
+        const designators = designators_data[0..header.counts.charcnt];
+        if (designators[designators.len - 1] != 0) return error.Malformed; // rfc8536: charcnt [...] includes the trailing NUL (0x00) octet
+
+        // Iterate through the timetypes again, setting the designator names
+        for (timetypes) |*tt| {
+            const name = std.mem.sliceTo(designators[tt.name_data[0]..], 0);
+            // We are mandating the "SHOULD" 6-character limit so we can pack the struct better, and to conform to POSIX.
+            if (name.len > 6) return error.Malformed; // rfc8536: Time zone designations SHOULD consist of at least three (3) and no more than six (6) ASCII characters.
+            std.mem.copy(u8, tt.name_data[0..], name);
+            tt.name_data[name.len] = 0;
+        }
+
+        // Parse leap seconds
+        i = 0;
+        while (i < header.counts.leapcnt) : (i += 1) {
+            const occur: i64 = if (legacy) try reader.readIntBig(i32) else try reader.readIntBig(i64);
+            if (occur < 0) return error.Malformed; // rfc8536: occur [...] MUST be nonnegative
+            if (i > 0 and leapseconds[i - 1].occurrence + 2419199 > occur) return error.Malformed; // rfc8536: occur [...] each later value MUST be at least 2419199 greater than the previous value
+            if (occur > std.math.maxInt(i48)) return error.Malformed; // Unreasonably far into the future
+
+            const corr = try reader.readIntBig(i32);
+            if (i == 0 and corr != -1 and corr != 1) return error.Malformed; // rfc8536: The correction value in the first leap-second record, if present, MUST be either one (1) or minus one (-1)
+            if (i > 0 and leapseconds[i - 1].correction != corr + 1 and leapseconds[i - 1].correction != corr - 1) return error.Malformed; // rfc8536: The correction values in adjacent leap-second records MUST differ by exactly one (1)
+            if (corr > std.math.maxInt(i16)) return error.Malformed; // Unreasonably large correction
+
+            leapseconds[i] = .{
+                .occurrence = @intCast(i48, occur),
+                .correction = @intCast(i16, corr),
+            };
+        }
+
+        // Parse standard/wall indicators
+        i = 0;
+        while (i < header.counts.isstdcnt) : (i += 1) {
+            const stdtime = try reader.readByte();
+            if (stdtime == 1) {
+                timetypes[i].flags |= 0x02;
+            }
+        }
+
+        // Parse UT/local indicators
+        i = 0;
+        while (i < header.counts.isutcnt) : (i += 1) {
+            const ut = try reader.readByte();
+            if (ut == 1) {
+                timetypes[i].flags |= 0x04;
+                if (!timetypes[i].standardTimeIndicator()) return error.Malformed; // rfc8536: standard/wall value MUST be one (1) if the UT/local value is one (1)
+            }
+        }
+
+        // Footer
+        var footer: ?[]u8 = null;
+        if (!legacy) {
+            if ((try reader.readByte()) != '\n') return error.Malformed; // An rfc8536 footer must start with a newline
+            var footerdata_buf: [128]u8 = undefined;
+            const footer_mem = reader.readUntilDelimiter(&footerdata_buf, '\n') catch |err| switch (err) {
+                error.StreamTooLong => return error.OverlargeFooter, // Read more than 128 bytes, much larger than any reasonable POSIX TZ string
+                else => return err,
+            };
+            if (footer_mem.len != 0) {
+                footer = try allocator.dupe(u8, footer_mem);
+            }
+        }
+        errdefer if (footer) |ft| allocator.free(ft);
+
+        return Tz{
+            .allocator = allocator,
+            .transitions = transitions,
+            .timetypes = timetypes,
+            .leapseconds = leapseconds,
+            .footer = footer,
+        };
+    }
+
+    pub fn deinit(self: *Tz) void {
+        if (self.footer) |footer| {
+            self.allocator.free(footer);
+        }
+        self.allocator.free(self.leapseconds);
+        self.allocator.free(self.transitions);
+        self.allocator.free(self.timetypes);
+    }
+};
+
+test "slim" {
+    const data = @embedFile("tz/asia_tokyo.tzif");
+    var in_stream = std.io.fixedBufferStream(data);
+
+    var tz = try std.Tz.parse(std.testing.allocator, in_stream.reader());
+    defer tz.deinit();
+
+    try std.testing.expectEqual(tz.transitions.len, 9);
+    try std.testing.expect(std.mem.eql(u8, tz.transitions[3].timetype.name(), "JDT"));
+    try std.testing.expectEqual(tz.transitions[5].ts, -620298000); // 1950-05-06 15:00:00 UTC
+    try std.testing.expectEqual(tz.leapseconds[13].occurrence, 567993613); // 1988-01-01 00:00:00 UTC (+23s in TAI, and +13 in the data since it doesn't store the initial 10 second offset)
+}
+
+test "fat" {
+    const data = @embedFile("tz/antarctica_davis.tzif");
+    var in_stream = std.io.fixedBufferStream(data);
+
+    var tz = try std.Tz.parse(std.testing.allocator, in_stream.reader());
+    defer tz.deinit();
+
+    try std.testing.expectEqual(tz.transitions.len, 8);
+    try std.testing.expect(std.mem.eql(u8, tz.transitions[3].timetype.name(), "+05"));
+    try std.testing.expectEqual(tz.transitions[4].ts, 1268251224); // 2010-03-10 20:00:00 UTC
+}
+
+test "legacy" {
+    // Taken from Slackware 8.0, from 2001
+    const data = @embedFile("tz/europe_vatican.tzif");
+    var in_stream = std.io.fixedBufferStream(data);
+
+    var tz = try std.Tz.parse(std.testing.allocator, in_stream.reader());
+    defer tz.deinit();
+
+    try std.testing.expectEqual(tz.transitions.len, 170);
+    try std.testing.expect(std.mem.eql(u8, tz.transitions[69].timetype.name(), "CET"));
+    try std.testing.expectEqual(tz.transitions[123].ts, 1414285200); // 2014-10-26 01:00:00 UTC
+}
diff --git a/lib/std/tz/antarctica_davis.tzif b/lib/std/tz/antarctica_davis.tzif
new file mode 100644
index 0000000000..662aa0098a
--- /dev/null
+++ b/lib/std/tz/antarctica_davis.tzif
diff --git a/lib/std/tz/asia_tokyo.tzif b/lib/std/tz/asia_tokyo.tzif
new file mode 100644
index 0000000000..15d780f7a8
--- /dev/null
+++ b/lib/std/tz/asia_tokyo.tzif
diff --git a/lib/std/tz/europe_vatican.tzif b/lib/std/tz/europe_vatican.tzif
new file mode 100644
index 0000000000..fe55064d1d
--- /dev/null
+++ b/lib/std/tz/europe_vatican.tzif