zig/lib/std/io/reader.zig

355 lines
14 KiB
Zig

// SPDX-License-Identifier: MIT
// Copyright (c) 2015-2021 Zig Contributors
// This file is part of [zig](https://ziglang.org/), which is MIT licensed.
// The MIT license requires this copyright notice to be included in all copies
// and substantial portions of the software.
const std = @import("../std.zig");
const builtin = std.builtin;
const math = std.math;
const assert = std.debug.assert;
const mem = std.mem;
const testing = std.testing;
pub fn Reader(
comptime Context: type,
comptime ReadError: type,
/// Returns the number of bytes read. It may be less than buffer.len.
/// If the number of bytes read is 0, it means end of stream.
/// End of stream is not an error condition.
comptime readFn: fn (context: Context, buffer: []u8) ReadError!usize,
) type {
return struct {
pub const Error = ReadError;
context: Context,
const Self = @This();
/// Returns the number of bytes read. It may be less than buffer.len.
/// If the number of bytes read is 0, it means end of stream.
/// End of stream is not an error condition.
pub fn read(self: Self, buffer: []u8) Error!usize {
return readFn(self.context, buffer);
}
/// Returns the number of bytes read. If the number read is smaller than `buffer.len`, it
/// means the stream reached the end. Reaching the end of a stream is not an error
/// condition.
pub fn readAll(self: Self, buffer: []u8) Error!usize {
var index: usize = 0;
while (index != buffer.len) {
const amt = try self.read(buffer[index..]);
if (amt == 0) return index;
index += amt;
}
return index;
}
/// If the number read would be smaller than `buf.len`, `error.EndOfStream` is returned instead.
pub fn readNoEof(self: Self, buf: []u8) !void {
const amt_read = try self.readAll(buf);
if (amt_read < buf.len) return error.EndOfStream;
}
pub const readAllBuffer = @compileError("deprecated; use readAllArrayList()");
/// Appends to the `std.ArrayList` contents by reading from the stream until end of stream is found.
/// If the number of bytes appended would exceed `max_append_size`, `error.StreamTooLong` is returned
/// and the `std.ArrayList` has exactly `max_append_size` bytes appended.
pub fn readAllArrayList(self: Self, array_list: *std.ArrayList(u8), max_append_size: usize) !void {
return self.readAllArrayListAligned(null, array_list, max_append_size);
}
pub fn readAllArrayListAligned(
self: Self,
comptime alignment: ?u29,
array_list: *std.ArrayListAligned(u8, alignment),
max_append_size: usize,
) !void {
try array_list.ensureCapacity(math.min(max_append_size, 4096));
const original_len = array_list.items.len;
var start_index: usize = original_len;
while (true) {
array_list.expandToCapacity();
const dest_slice = array_list.items[start_index..];
const bytes_read = try self.readAll(dest_slice);
start_index += bytes_read;
if (start_index - original_len > max_append_size) {
array_list.shrinkAndFree(original_len + max_append_size);
return error.StreamTooLong;
}
if (bytes_read != dest_slice.len) {
array_list.shrinkAndFree(start_index);
return;
}
// This will trigger ArrayList to expand superlinearly at whatever its growth rate is.
try array_list.ensureCapacity(start_index + 1);
}
}
/// Allocates enough memory to hold all the contents of the stream. If the allocated
/// memory would be greater than `max_size`, returns `error.StreamTooLong`.
/// Caller owns returned memory.
/// If this function returns an error, the contents from the stream read so far are lost.
pub fn readAllAlloc(self: Self, allocator: *mem.Allocator, max_size: usize) ![]u8 {
var array_list = std.ArrayList(u8).init(allocator);
defer array_list.deinit();
try self.readAllArrayList(&array_list, max_size);
return array_list.toOwnedSlice();
}
/// Replaces the `std.ArrayList` contents by reading from the stream until `delimiter` is found.
/// Does not include the delimiter in the result.
/// If the `std.ArrayList` length would exceed `max_size`, `error.StreamTooLong` is returned and the
/// `std.ArrayList` is populated with `max_size` bytes from the stream.
pub fn readUntilDelimiterArrayList(
self: Self,
array_list: *std.ArrayList(u8),
delimiter: u8,
max_size: usize,
) !void {
array_list.shrinkRetainingCapacity(0);
while (true) {
var byte: u8 = try self.readByte();
if (byte == delimiter) {
return;
}
if (array_list.items.len == max_size) {
return error.StreamTooLong;
}
try array_list.append(byte);
}
}
/// Allocates enough memory to read until `delimiter`. If the allocated
/// memory would be greater than `max_size`, returns `error.StreamTooLong`.
/// Caller owns returned memory.
/// If this function returns an error, the contents from the stream read so far are lost.
pub fn readUntilDelimiterAlloc(
self: Self,
allocator: *mem.Allocator,
delimiter: u8,
max_size: usize,
) ![]u8 {
var array_list = std.ArrayList(u8).init(allocator);
defer array_list.deinit();
try self.readUntilDelimiterArrayList(&array_list, delimiter, max_size);
return array_list.toOwnedSlice();
}
/// Allocates enough memory to read until `delimiter` or end-of-stream.
/// If the allocated memory would be greater than `max_size`, returns
/// `error.StreamTooLong`. If end-of-stream is found, returns the rest
/// of the stream. If this function is called again after that, returns
/// null.
/// Caller owns returned memory.
/// If this function returns an error, the contents from the stream read so far are lost.
pub fn readUntilDelimiterOrEofAlloc(
self: Self,
allocator: *mem.Allocator,
delimiter: u8,
max_size: usize,
) !?[]u8 {
var array_list = std.ArrayList(u8).init(allocator);
defer array_list.deinit();
self.readUntilDelimiterArrayList(&array_list, delimiter, max_size) catch |err| switch (err) {
error.EndOfStream => if (array_list.items.len == 0) {
return null;
} else {
return array_list.toOwnedSlice();
},
else => |e| return e,
};
return array_list.toOwnedSlice();
}
/// Reads from the stream until specified byte is found. If the buffer is not
/// large enough to hold the entire contents, `error.StreamTooLong` is returned.
/// If end-of-stream is found, returns the rest of the stream. If this
/// function is called again after that, returns null.
/// Returns a slice of the stream data, with ptr equal to `buf.ptr`. The
/// delimiter byte is not included in the returned slice.
pub fn readUntilDelimiterOrEof(self: Self, buf: []u8, delimiter: u8) !?[]u8 {
var index: usize = 0;
while (true) {
const byte = self.readByte() catch |err| switch (err) {
error.EndOfStream => {
if (index == 0) {
return null;
} else {
return buf[0..index];
}
},
else => |e| return e,
};
if (byte == delimiter) return buf[0..index];
if (index >= buf.len) return error.StreamTooLong;
buf[index] = byte;
index += 1;
}
}
/// Reads from the stream until specified byte is found, discarding all data,
/// including the delimiter.
/// If end-of-stream is found, this function succeeds.
pub fn skipUntilDelimiterOrEof(self: Self, delimiter: u8) !void {
while (true) {
const byte = self.readByte() catch |err| switch (err) {
error.EndOfStream => return,
else => |e| return e,
};
if (byte == delimiter) return;
}
}
/// Reads 1 byte from the stream or returns `error.EndOfStream`.
pub fn readByte(self: Self) !u8 {
var result: [1]u8 = undefined;
const amt_read = try self.read(result[0..]);
if (amt_read < 1) return error.EndOfStream;
return result[0];
}
/// Same as `readByte` except the returned byte is signed.
pub fn readByteSigned(self: Self) !i8 {
return @bitCast(i8, try self.readByte());
}
/// Reads exactly `num_bytes` bytes and returns as an array.
/// `num_bytes` must be comptime-known
pub fn readBytesNoEof(self: Self, comptime num_bytes: usize) ![num_bytes]u8 {
var bytes: [num_bytes]u8 = undefined;
try self.readNoEof(&bytes);
return bytes;
}
/// Reads a native-endian integer
pub fn readIntNative(self: Self, comptime T: type) !T {
const bytes = try self.readBytesNoEof((@typeInfo(T).Int.bits + 7) / 8);
return mem.readIntNative(T, &bytes);
}
/// Reads a foreign-endian integer
pub fn readIntForeign(self: Self, comptime T: type) !T {
const bytes = try self.readBytesNoEof((@typeInfo(T).Int.bits + 7) / 8);
return mem.readIntForeign(T, &bytes);
}
pub fn readIntLittle(self: Self, comptime T: type) !T {
const bytes = try self.readBytesNoEof((@typeInfo(T).Int.bits + 7) / 8);
return mem.readIntLittle(T, &bytes);
}
pub fn readIntBig(self: Self, comptime T: type) !T {
const bytes = try self.readBytesNoEof((@typeInfo(T).Int.bits + 7) / 8);
return mem.readIntBig(T, &bytes);
}
pub fn readInt(self: Self, comptime T: type, endian: builtin.Endian) !T {
const bytes = try self.readBytesNoEof((@typeInfo(T).Int.bits + 7) / 8);
return mem.readInt(T, &bytes, endian);
}
pub fn readVarInt(self: Self, comptime ReturnType: type, endian: builtin.Endian, size: usize) !ReturnType {
assert(size <= @sizeOf(ReturnType));
var bytes_buf: [@sizeOf(ReturnType)]u8 = undefined;
const bytes = bytes_buf[0..size];
try self.readNoEof(bytes);
return mem.readVarInt(ReturnType, bytes, endian);
}
/// Optional parameters for `skipBytes`
pub const SkipBytesOptions = struct {
buf_size: usize = 512,
};
// `num_bytes` is a `u64` to match `off_t`
/// Reads `num_bytes` bytes from the stream and discards them
pub fn skipBytes(self: Self, num_bytes: u64, comptime options: SkipBytesOptions) !void {
var buf: [options.buf_size]u8 = undefined;
var remaining = num_bytes;
while (remaining > 0) {
const amt = std.math.min(remaining, options.buf_size);
try self.readNoEof(buf[0..amt]);
remaining -= amt;
}
}
/// Reads `slice.len` bytes from the stream and returns if they are the same as the passed slice
pub fn isBytes(self: Self, slice: []const u8) !bool {
var i: usize = 0;
var matches = true;
while (i < slice.len) : (i += 1) {
if (slice[i] != try self.readByte()) {
matches = false;
}
}
return matches;
}
pub fn readStruct(self: Self, comptime T: type) !T {
// Only extern and packed structs have defined in-memory layout.
comptime assert(@typeInfo(T).Struct.layout != builtin.TypeInfo.ContainerLayout.Auto);
var res: [1]T = undefined;
try self.readNoEof(mem.sliceAsBytes(res[0..]));
return res[0];
}
/// Reads an integer with the same size as the given enum's tag type. If the integer matches
/// an enum tag, casts the integer to the enum tag and returns it. Otherwise, returns an error.
/// TODO optimization taking advantage of most fields being in order
pub fn readEnum(self: Self, comptime Enum: type, endian: builtin.Endian) !Enum {
const E = error{
/// An integer was read, but it did not match any of the tags in the supplied enum.
InvalidValue,
};
const type_info = @typeInfo(Enum).Enum;
const tag = try self.readInt(type_info.tag_type, endian);
inline for (std.meta.fields(Enum)) |field| {
if (tag == field.value) {
return @field(Enum, field.name);
}
}
return E.InvalidValue;
}
};
}
test "Reader" {
var buf = "a\x02".*;
const reader = std.io.fixedBufferStream(&buf).reader();
testing.expect((try reader.readByte()) == 'a');
testing.expect((try reader.readEnum(enum(u8) {
a = 0,
b = 99,
c = 2,
d = 3,
}, undefined)) == .c);
testing.expectError(error.EndOfStream, reader.readByte());
}
test "Reader.isBytes" {
const reader = std.io.fixedBufferStream("foobar").reader();
testing.expectEqual(true, try reader.isBytes("foo"));
testing.expectEqual(false, try reader.isBytes("qux"));
}
test "Reader.skipBytes" {
const reader = std.io.fixedBufferStream("foobar").reader();
try reader.skipBytes(3, .{});
testing.expect(try reader.isBytes("bar"));
try reader.skipBytes(0, .{});
testing.expectError(error.EndOfStream, reader.skipBytes(1, .{}));
}