mirrors/zig
1
0
Fork 0
mirror of https://codeberg.org/ziglang/zig.git synced 2025-12-06 05:44:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

tz parsing reader interface, test thicc files, and exclude tzif

Browse source
This commit is contained in:
Jens Goldberg 2021-12-31 17:17:49 +00:00
parent e4672c95f1
commit 9a56435666
4 changed files with 106 additions and 89 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
build.zig
View file

@ -98,6 +98,7 @@ pub fn build(b: *Builder) !void {
".z.9",
".gz",
"rfc1951.txt",
".tzif",
},
.blank_extensions = &[_][]const u8{
"test.zig",

2
lib/std/std.zig
View file

@ -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");
@ -81,7 +82,6 @@ pub const sort = @import("sort.zig");
pub const ascii = @import("ascii.zig");
pub const testing = @import("testing.zig");
pub const time = @import("time.zig");
pub const tz = @import("tz.zig");
pub const unicode = @import("unicode.zig");
pub const valgrind = @import("valgrind.zig");
pub const wasm = @import("wasm.zig");

192
lib/std/tz.zig
View file

@ -40,160 +40,163 @@ pub const Tz = struct {
leapseconds: []const Leapsecond,
footer: []const u8,
pub fn parse(allocator: std.mem.Allocator, data: []const u8) !Tz {
const header_size = 4 + 1 + 15 + 6 * 4;
if (data.len < header_size) return error.BadSize;
pub fn parse(allocator: std.mem.Allocator, reader: anytype) !Tz {
_ = allocator;
const Header = extern struct {
magic: [4]u8,
version: u8,
reserved: [15]u8,
};
const magic_l = data[0..4];
const version_l = data[4];
if (!std.mem.eql(u8, magic_l, "TZif")) return error.BadHeader;
if (version_l != '2' and version_l != '3') return error.BadVersion;
const Counts = extern struct {
isutcnt: u32,
isstdcnt: u32,
leapcnt: u32,
timecnt: u32,
typecnt: u32,
charcnt: u32,
};
// Parse the legacy header and skip the entire thing
const isutcnt_l = std.mem.readIntBig(u32, data[20..24]);
const isstdcnt_l = std.mem.readIntBig(u32, data[24..28]);
const leapcnt_l = std.mem.readIntBig(u32, data[28..32]);
const timecnt_l = std.mem.readIntBig(u32, data[32..36]);
const typecnt_l = std.mem.readIntBig(u32, data[36..40]);
const charcnt_l = std.mem.readIntBig(u32, data[40..44]);
const data_block_size_legacy = timecnt_l * 5 + typecnt_l * 6 + charcnt_l + leapcnt_l * 8 + isstdcnt_l + isutcnt_l;
if (data.len < header_size + data_block_size_legacy) return error.BadSize;
// Parse and skip the legacy header and data
{
const header = try reader.readStruct(Header);
if (!std.mem.eql(u8, &header.magic, "TZif")) return error.BadHeader;
if (header.version == 0) return error.UnsupportedLegacyFormat;
if (header.version != '2' and header.version != '3') return error.BadVersion;
const data2 = data[header_size + data_block_size_legacy ..];
if (data2.len < header_size) return error.BadSize;
var counts = try reader.readStruct(Counts);
if (builtin.target.cpu.arch.endian() != std.builtin.Endian.Big) {
std.mem.bswapAllFields(Counts, &counts);
}
const magic = data2[0..4];
const version = data2[4];
if (!std.mem.eql(u8, magic, "TZif")) return error.BadHeader;
if (version != '2' and version != '3') return error.BadVersion;
const skipv = counts.timecnt * 5 + counts.typecnt * 6 + counts.charcnt + counts.leapcnt * 8 + counts.isstdcnt + counts.isutcnt;
try reader.skipBytes(skipv, .{});
}
const isutcnt = std.mem.readIntBig(u32, data2[20..24]);
const isstdcnt = std.mem.readIntBig(u32, data2[24..28]);
const leapcnt = std.mem.readIntBig(u32, data2[28..32]);
const timecnt = std.mem.readIntBig(u32, data2[32..36]);
const typecnt = std.mem.readIntBig(u32, data2[36..40]);
const charcnt = std.mem.readIntBig(u32, data2[40..44]);
const 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 (isstdcnt != 0 and isstdcnt != typecnt) return error.Malformed; // rfc8536: isstdcnt [...] MUST either be zero or equal to "typecnt"
if (isutcnt != 0 and isutcnt != typecnt) return error.Malformed; // rfc8536: isutcnt [...] MUST either be zero or equal to "typecnt"
if (typecnt == 0) return error.Malformed; // rfc8536: typecnt [...] MUST NOT be zero
if (charcnt == 0) return error.Malformed; // rfc8536: charcnt [...] MUST NOT be zero
var counts = try reader.readStruct(Counts);
if (builtin.target.cpu.arch.endian() != std.builtin.Endian.Big) {
std.mem.bswapAllFields(Counts, &counts);
}
const data_block_size = timecnt * 9 + typecnt * 6 + charcnt + leapcnt * 12 + isstdcnt + isutcnt;
if (data2.len < header_size + data_block_size) return error.BadSize;
if (counts.isstdcnt != 0 and counts.isstdcnt != counts.typecnt) return error.Malformed; // rfc8536: isstdcnt [...] MUST either be zero or equal to "typecnt"
if (counts.isutcnt != 0 and counts.isutcnt != counts.typecnt) return error.Malformed; // rfc8536: isutcnt [...] MUST either be zero or equal to "typecnt"
if (counts.typecnt == 0) return error.Malformed; // rfc8536: typecnt [...] MUST NOT be zero
if (counts.charcnt == 0) return error.Malformed; // rfc8536: charcnt [...] MUST NOT be zero
if (counts.charcnt > 256 + 6) return error.Malformed; // Not explicitly banned by rfc8536 but nonsensical
var leapseconds = try allocator.alloc(Leapsecond, leapcnt);
var leapseconds = try allocator.alloc(Leapsecond, counts.leapcnt);
errdefer allocator.free(leapseconds);
var transitions = try allocator.alloc(Transition, timecnt);
var transitions = try allocator.alloc(Transition, counts.timecnt);
errdefer allocator.free(transitions);
var timetypes = try allocator.alloc(Timetype, typecnt);
var timetypes = try allocator.alloc(Timetype, counts.typecnt);
errdefer allocator.free(timetypes);
var p: usize = header_size;
// First, parse timezone designators ahead of time so that we can reject malformed files early
const designators = data2[header_size + timecnt * 9 + typecnt * 6 .. header_size + timecnt * 9 + typecnt * 6 + charcnt];
if (designators[designators.len - 1] != 0) return error.Malformed; // rfc8536: charcnt [...] includes the trailing NUL (0x00) octet
// Parse transition types
var i: usize = 0;
while (i < timecnt) : (i += 1) {
transitions[i].ts = std.mem.readIntSliceBig(i64, data2[p .. p + 8]);
p += 8;
while (i < counts.timecnt) : (i += 1) {
transitions[i].ts = try reader.readIntBig(i64);
}
i = 0;
while (i < timecnt) : (i += 1) {
const tt = data2[p];
while (i < 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];
p += 1;
}
// Parse time types
i = 0;
while (i < typecnt) : (i += 1) {
const offset = std.mem.readIntSliceBig(i32, data2[p .. p + 4]);
while (i < 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 = data2[p + 4];
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 = data2[p + 5];
if (idx > designators.len - 1) return error.Malformed; // rfc8536: (desig)idx [...] Each index MUST be in the range [0, "charcnt" - 1]
const name = std.mem.sliceTo(designators[idx..], 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.
const idx = try reader.readByte();
if (idx > 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,
};
std.mem.copy(u8, timetypes[i].name_data[0..], name);
timetypes[i].name_data[name.len] = 0;
p += 6;
// Temporarily cache idx in name_data to be processed after we've read the designator names below
timetypes[i].name_data[0] = idx;
}
// Skip the designators we got earlier
p += charcnt;
var designators_data: [256 + 6]u8 = undefined;
try reader.readNoEof(designators_data[0..counts.charcnt]);
const designators = designators_data[0..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 < leapcnt) : (i += 1) {
const occur = std.mem.readIntSliceBig(i64, data2[p .. p + 8]);
while (i < counts.leapcnt) : (i += 1) {
const occur = 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 = std.mem.readIntSliceBig(i32, data2[p + 8 .. p + 12]);
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 (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),
};
p += 12;
}
// Parse standard/wall indicators
i = 0;
while (i < isstdcnt) : (i += 1) {
const stdtime = data2[p];
while (i < counts.isstdcnt) : (i += 1) {
const stdtime = try reader.readByte();
if (stdtime == 1) {
timetypes[i].flags |= 0x02;
}
p += 1;
}
// Parse UT/local indicators
i = 0;
while (i < isutcnt) : (i += 1) {
const ut = data2[p];
while (i < 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)
}
p += 1;
}
if ((try reader.readByte()) != '\n') return error.Malformed; // An rfc8536 footer must start with a newline
// Footer
if (data2[p..].len < 2) return error.Malformed; // rfc8536 requires at least 2 newlines
if (data2[p] != '\n') return error.Malformed; // Not a rfc8536 footer
const footer_end = std.mem.indexOfScalar(u8, data2[p + 1 ..], '\n') orelse return error.Malformed; // No 2nd rfc8536 newline
const footer = try allocator.dupe(u8, data2[p + 1 .. p + 1 + footer_end]);
errdefer allocator.free(footer);
var footerdata_buf: [128]u8 = undefined;
const footer = 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,
};
const footer_dup = try allocator.dupe(u8, footer);
errdefer allocator.free(footer_dup);
return Tz{
.allocator = allocator,
.transitions = transitions,
.timetypes = timetypes,
.leapseconds = leapseconds,
.footer = footer,
.footer = footer_dup,
};
}
@ -205,14 +208,27 @@ pub const Tz = struct {
}
};
test "parse" {
// Asia/Tokyo is good for embedding, as Japan only had DST for a short while during the US occupation
test "slim" {
const data = @embedFile("tz/asia_tokyo.tzif");
var tz = try Tz.parse(std.testing.allocator, data);
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 00:00:13 (IAT)
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
}

BIN
lib/std/tz/antarctica_davis.tzif Normal file
View file

Binary file not shown.