0.4.0-pre.1

UNRELEASED: 0.4.0

How to migrate your old codebase to 0.4.0-native architecture?

Compatibility Layer

To enable it, enable compat-v0_3 feature.

The built-in compatibility layer emulates what
v0.3.x used to behave, including the serialization,
deserialization, number encoding and number decoding.

With this emulation, you can migrate your code with less pain–
most of the current components will mostly behave as before. However,
all of them are marked as deprecated, and we may remove
this layer in the future major version (such as 0.5.x).

DO NOT DEPEND ON THIS LAYER, and migrate your codebase as long as possible! This layer may slow down your program, and the layer may have some unexpected changes compared to the 0.3.x version (for example, the Error variant).

Encoding (Now called Serializing)

Assuming your code is using encode_bbo (or encode_orderbook, whatever):

let encoded: Vec<u8> = encode_bbo(your_bbo_msg).unwrap();

First, you have to convert your Msg in crypto-crawler,
to libstock's Structure. You can approach it with try_into():

// OrderBookMsg → OrderbookStructure, KlineMsg → KlineStructure,
// TradeMsg → TradeStructure, FundingRateMsg → FundingRateStructure,
// so on.
let bbo_structure = BboStructure::try_from(your_bbo_msg)?

Our new serialization method accepts any type implementing Write –
in other words, you can pass your Socket, File and anything else writable
to the serialize() method:

// Use our serialization method.
use wmjtyd_libstock::data::serializer::StructureSerializer;

bbo_structure.serialize(your_file)?;

If you prefer the current way, you need to manage your own buffer
(however, you can pass any buffer implementing Write for your case.)

let mut buffer = Vec::new();
bbo_structure.serialize(buffer)?;

Decoding (Now called Deserializing)

Assuming your code is using decode_bbo (or decode_orderbook, whatever):

let decoded: BboMsg = decode_bbo(src).unwrap();

First, use our deserialization method to construct the Structure.
It accepts any type implementing Read trait, so you can just
pass your socket, file or anything else readable to the deserialize() method.

// Use our deserialization method.
use wmjtyd_libstock::data::serializer::StructureDeserializer;

let decoded = BboStructure::deserialize(&mut your_data_source);

Note that if your data source is a Vec<T>, you must turn it to a mutable-immutable slice.
It is counterintuitive, but it works ;)

let decoded = BboStructure::deserialize(&mut your_vec.as_slice());

Finally, convert your Structure back to the crypto-crawler's Msg:

let msg = BboMsg::try_from(decoded)?;

Number Encoding and Decoding

We have rewritten the encoding and decoding methods for clearer
architecture, robust and performance. However, it introduces
some breaking change though it can be easily resolved.

Assuming you have such this code:

let num = 1234_5678.to_string();
u32::encode_bytes(&num)

First, You need to convert your integer, float number, or number string
to rust_decimal::Decimal. We have re-exported it in data::num
so you won't need to add another dependencies by yourself.

Take the above as an example, we can rewrite it to:

// Integer
let num = rust_decimal::from(1234_5678);
// Float; you should do your own error handling.
let num = rust_decimal::from_f32(1234.5678).map_err(...)?;
// Number string; you should do your own error handling.
let num = rust_decimal::from_str_exact("1234.5678")?;

After that, we include our new encode API and encode the num:

use wmjtyd_libstock::data::num::Encoder;

let num_bytes: [u8; 5] = num.encode()?;

To decode the number, use Decoder::decode:

use wmjtyd_libstock::data::num::Decoder;

let num = Decoder::decode(&num_bytes)?;

Fields

We did not write the compatibility layer for fields,
as it used to be internally used in the serialization and deserialization, and most people may not operate with this.

Some notable big changes to migrate your codebase:

• data::types has merged into data::fields.
• All the -Repr are renamed to -Field.
• try_from_reader → deserialize_from_reader
• from_bytes (or try_from_bytes) → deserialize
• to_bytes (or try_to_bytes) → serialize
• SymbolPairField is no longer a tuple – its fields has been named now.
• ReadExt has been merged to Deserializer.

message

Though we reshaped and rewrote all the message modules,
there are not really big changes for users.

Migrate nanomsg writers

Assuming you used to write this way:

use std::io::Write;
use wmjtyd_libstock::message::nanomsg::{Nanomsg, NanomsgProtocol};

let nanomsg = Nanomsg::new("ipc:///tmp/cl-nanomsg-old-api-w.ipc", NanomsgProtocol::Pub);

if let Ok(mut nanomsg) = nanomsg {
    nanomsg.write_all(b"Hello World!").ok();
}

First, update your Nanomsg imports to NanomsgPublisher.
We have separated Nanomsg to NanomsgPublisher and NanomsgSubscriber,
which implementing the corresponding Publisher and Subscriber:

use wmjtyd_libstock::message::nanomsg::NanomsgPublisher;

And then, construct your NanomsgPublisher. Note that no any
additional parameter need to pass in this construction. For address,
we will bind our address in another function later.

let nanomsg = NanomsgPublisher::new();

new() returns Result, so error handling is needed here.
After that, let's bind our address to Nanomsg:

use wmjtyd_libstock::message::traits::Bind;

if let Ok(mut nanomsg) = nanomsg {
    nanomsg.bind("ipc:///tmp/cl-nanomsg-new-api-w.ipc").ok();

Note that we includes traits::Bind here. To use the bind
method, you must introduce our Bind trait here. It is
necessary for our high-level abstraction. Finally, we introduce
Write trait, just like what we did before.

use std::io::Write;  // or use wmjtyd_libstock::message::traits::Write;
    nanomsg.write_all(b"Hello World!").ok();
}

Besides, we provide the async API of nanomsg now!

use tokio::io::AsyncWriteExt;  // or use wmjtyd_libstock::message::traits::AsyncWriteExt;
    nanomsg.write_all(b"Hello World!").await.ok();
}

The full synchronous example:

use wmjtyd_libstock::message::nanomsg::NanomsgPublisher;
use wmjtyd_libstock::message::traits::{Bind, Write};

let nanomsg = NanomsgPublisher::new();

if let Ok(mut nanomsg) = nanomsg {
    nanomsg.bind("ipc:///tmp/cl-nanomsg-new-api-w.ipc").ok();
    nanomsg.write_all(b"Hello World!").ok();
}

Migrate zeromq writers

Just like nanomsg, but use zeromq::ZeromqPublisher. Asynchronous Example:

use wmjtyd_libstock::message::traits::{AsyncWriteExt, Bind};
use wmjtyd_libstock::message::zeromq::ZeromqPublisher;

let zeromq = ZeromqPublisher::new();

if let Ok(mut zeromq) = zeromq {
    zeromq.bind("ipc:///tmp/cl-zeromq-new-api-w-a.ipc").ok();
    zeromq.write_all(b"Hello World!").await.ok();
}

Migrate nanomsg readers

WIP. Really similar to what we did in writers.
Example of this new synchronous API:

use wmjtyd_libstock::message::nanomsg::NanomsgSubscriber;
use wmjtyd_libstock::message::traits::{Connect, Read, Subscribe};

let nanomsg = NanomsgSubscriber::new();

if let Ok(mut nanomsg) = nanomsg {
    nanomsg.connect("ipc:///tmp/cl-nanomsg-new-api-r.ipc").ok();
    nanomsg.subscribe(b"").ok();

    let mut buf = [0; 12];
    nanomsg.read_exact(&mut buf).ok();
    assert_eq!(b"Hello World!", &buf);
}

Migrate zeromq readers

WIP. All of the APIs are followed our new rule and
thus the logic is just the same. Example of this new
asynchronous API:

use wmjtyd_libstock::message::zeromq::ZeromqSubscriber;
use wmjtyd_libstock::message::traits::{Connect, AsyncReadExt, Subscribe};

let zeromq = ZeromqSubscriber::new();

if let Ok(mut zeromq) = zeromq {
    zeromq.connect("ipc:///tmp/cl-zeromq-new-api-r.ipc").ok();
    zeromq.subscribe(b"").ok();

    let mut buf = [0; 12];
    zeromq.read_exact(&mut buf).await.ok();
    assert_eq!(b"Hello World!", &buf);
}

0.4.0 – Breaking Changes

• message has been largely rewritten.
• Methods under data module are mostly reshaped. You may need to
  rewrite your code to adapt this new architectiure.
  • We have abstracted a Structure – To serialize your
    BboMsg, do .try_into() first.
• Added a End-Of-Data flag (\0) to all the current structure.

0.4.0 – Features

• Rewrite data module for robust and performance.
• Implement Eq for file/writer.
• Implement a general serializer and deserializer for field and structure.
  • StructSerializer: We now accept any type implemented std::io::Write,
    instead of BufWriter only!
  • StructDeserializer: We now accept any type implemented std::io::Read,
    instead of BufReader only!
• Add the ability to convert other structures other than what
  crypto-crawler provides! 🎉
  • Just implement the TryFrom between your structure and our structure.
  • Our incoming C binding adds the ability to operate with libstock's
    serialization and deserialization feature in C/C++.
  • Our work-in-progress Java binding also adds such the ability
    like C/C++'s.
• Derive more useful traits for our structures, such as Debug, PartialEq,
  Eq, Hash, Clone.
  • Also our fields ;)

0.4.0 – Chores

• Update dependencies.
  • WIP: merge crypto-crawler changes to upstream

0.4.0 – CI

• Automatically abort the old CI tasks by
  setting the concurrent flag....

0.3.0-beta.3

Unreleased: 0.3.0

Unreleased: 0.3.0 – Breaking changes

• file/reader: open() returns Result instead of Option now.
  • For better backtrace
• file/writer: Set the size of the length field of written data structure to 2 bytes.
  • It used to be 'usize'. It is an implementation mistake.
• file/hex: removed two deprecated methods
• file/reader: move path to <date>/<name>.csv
• Updated the fields of data.
• file/datadir: need to pass a timestamp

Unreleased: 0.3.0 – Features

• slack: For sending notifications to Slack with Slack Hook.
• message: A basic encap of nanomsg and zeromq for subscribing and publishing.
  • (beta 2) implement DerefMut to Socket for message
• message: Add Subscribe trait for generalize subscribe() method.
  • zeromq and nanomsg both have implemented this.
• file/hex: Add the encode/decode capability for negative numbers
  • WIP: currently dirty but works.
• message/zeromq: migrate to the ZeroMQ implemented in Rust.

Unreleased: 0.3.0 – Bug fixes

• file/reader: Make read() returns the exact data
  • Currently, it always returns [] due to an implementation mistake.
• file/writer: Flush buffer after the data written ended.
• (beta 2) message: use connect instead of bind for Sub
• data/*/decode: ReceivedTimestampRepr should be 6 bytes instead of 8 bytes.

Unreleased: 0.3.0 – Refactoring

• Upgrade crypto-crawler-rs to 92aee0d37e228e53dd994a17058a7f819e005446
  • Clean up the unnecessary dependencies.
• Deprecated file/ident.

Unreleased: 0.3.0 – Tests

• Add encode/decode tests to data/bbo
  • WIP: add other encode/decode tests to data/*!

Unreleased: 0.3.0 – Chores

• file/reader: add info logger on new() for better debugging

Unreleased: 0.3.0 – CI

• Install nanomsg for message test.
• Introduce cargo nextest for faster build.

0.3.0-beta.2

Unreleased: 0.3.0

Unreleased: 0.3.0 – Breaking changes

• file/reader: open() returns Result instead of Option now.
  • For better backtrace
• file/writer: Set the size of the length field of written data structure to 2 bytes.
  • It used to be 'usize'. It is an implementation mistake.

Unreleased: 0.3.0 – Features

• slack: For sending notifications to Slack with Slack Hook.
• message: A basic encap of nanomsg and zeromq for subscribing and publishing.
  • (beta 2) implement DerefMut to Socket for message
• message: Add Subscribe trait for generalize subscribe() method.
  • zeromq and nanomsg both have implemented this.

Unreleased: 0.3.0 – Bug fixes

• file/reader: Make read() returns the exact data
  • Currently, it always returns [] due to an implementation mistake.
• file/writer: Flush buffer after the data written ended.
• (beta 2) message: use connect instead of bind for Sub
• data/*/decode: ReceivedTimestampRepr should be 6 bytes instead of 8 bytes.

Unreleased: 0.3.0 – Refactoring

• Upgrade crypto-crawler-rs to 92aee0d37e228e53dd994a17058a7f819e005446
  • Clean up the unnecessary dependencies.

Unreleased: 0.3.0 – Tests

• Add encode/decode tests to data/bbo
  • WIP: add other encode/decode tests to data/*!

Unreleased: 0.3.0 – Chores

• file/reader: add info logger on new() for better debugging

Unreleased: 0.3.0 – CI

• Install nanomsg for message test.
• Enable vendored-zeromq in CI by default.
• Introduce cargo nextest for faster build.

v0.3.0-beta.1

libstock's changelog

Unreleased: 0.3.0

Unreleased: 0.3.0 – Breaking changes

• file/reader: open() returns Result instead of Option now.
  • For better backtrace
• file/writer: Set the size of the length field of written data structure to 2 bytes.
  • It used to be 'usize'. It is an implementation mistake.

Unreleased: 0.3.0 – Features

• slack: For sending notifications to Slack with Slack Hook.
• message: A basic encap of [Socket] for subscribing and publishing.

Unreleased: 0.3.0 – Bug fixes

• file/reader: Make read() returns the exact data
  • Currently, it always returns [] due to an implementation mistake.
• file/writer: Flush buffer after the data written ended.

Unreleased: 0.3.0 – Tests

• file: Add integration test for writer & reader

Unreleased: 0.3.0 – Chores

• Upgrade crypto-crawler-rs to d45528283b387bbe6c46507d7837bf1dd229ff6d
• file/reader: add info logger on new() for better debugging

Unreleased: 0.3.0 – CI

• Install nanomsg for message test.

0.2.1-beta.1

Unreleased: 0.2.1

Unreleased: 0.2.1 – Features

• slack: for sending notifications to Slack with Slack Hook.
• message: a basic encap of [Socket] for subscribing and publishing.

Unreleased: 0.2.1 – Chores

• Upgrade crypto-crawler-rs to d45528283b387bbe6c46507d7837bf1dd229ff6d

Unreleased: 0.2.1 – CI

• Install nanomsg for message test.

0.2.0

0.2.0

0.2.0 - Breaking changes

• Symble is now renamed to Symbol.
• Symbol field is now serialized as the big endian representation of u8.
• Map EXCHANGE and INFOTYPE are replaced to the enumeration Exchange and InfoType.
  • Use strum to provide the map-like function.
• Some methods in data::hex are removed.
  • Replace your current code to the native method (.to_be_bytes())
  • (beta 5) For UNIX timestamp, use unix_ms_to_six_byte_hex and six_byte_hex_to_unix_ms instead.
• Removed and added some new unused error variants.

0.2.0 - Features

• data::orderbook, data::trade: Use BufReader instead of our based-on-Atomic seek reader.
• data::orderbook, data::trade: Use data::fields to serialize & deserialize.
• Separate the serialization & deserialization of fields to data::fields
• Replace the usize converter of hex::encode_num_to_bytes to a robust implementation
  • Related: Some methods in data::hex are removed.
• data::types: Add Exchange and InfoType while removing EXCHANGE and INFOTYPE.
• data::orderbook: Improve the logic of generate_diff.
• data::hex: Deprecate old inextensible methods (encode_num_to_bytes & decode_num_to_bytes)
  • use u32::encode_bytes and u32::decode_bytes for 5-byte encoding & decoding
• data::hex: Add the u64 to 10-byte encode & decode support
  • use u64::encode_bytes and u64::decode_bytes for 10-byte encoding & decoding
• data::types: Add more exchanges, and add PERIOD.
• Add the encoder and decoder of bbo and kline.
• file::writer: Add tracing log.
• file::writer: Refactored for extensibility.
• Change to wmjtyd's crypto-crawler and crypto-msg-parser.
• (beta 5) Add methods to serialize and deserialize UNIX timestamp to 6-byte u64.

0.2.0 - CI

• Also test if cargo doc can generate correct documentation.
• Run cargo clippy and cargo fmt --check

0.2.0 - Examples

• Add a production example of file::writer.
• (beta 4) data::hex add the testcase of i8 → u8 and u8 → i8

0.2.0 - Bug fixes

• file::writer: Don't create the directory if it has been existed.
• file::writer: .stop() can't work properly.
• (beta 4) {InfoTypeExpr,ExchangeTypeRepr}::try_from_str should be from lowercase
  • Added the test case of this.
• (beta 5) UNIX timestamp should serialize and deserialize to 6-byte u64.

0.2.0-beta.5

0.2.0 (beta)

0.2.0 (beta) - Breaking changes

• Symble is now renamed to Symbol.
• Symbol field is now serialized as the big endian representation of u8.
• Map EXCHANGE and INFOTYPE are replaced to the enumeration Exchange and InfoType.
  • Use strum to provide the map-like function.
• Some methods in data::hex are removed.
  • Replace your current code to the native method (.to_be_bytes())
  • (beta 5) For UNIX timestamp, use unix_ms_to_six_byte_hex and six_byte_hex_to_unix_ms instead.
• Removed and added some new unused error variants.

0.2.0 (beta) - Features

• data::orderbook, data::trade: Use BufReader instead of our based-on-Atomic seek reader.
• data::orderbook, data::trade: Use data::fields to serialize & deserialize.
• Separate the serialization & deserialization of fields to data::fields
• Replace the usize converter of hex::encode_num_to_bytes to a robust implementation
  • Related: Some methods in data::hex are removed.
• data::types: Add Exchange and InfoType while removing EXCHANGE and INFOTYPE.
• data::orderbook: Improve the logic of generate_diff.
• data::hex: Deprecate old inextensible methods (encode_num_to_bytes & decode_num_to_bytes)
  • use u32::encode_bytes and u32::decode_bytes for 5-byte encoding & decoding
• data::hex: Add the u64 to 10-byte encode & decode support
  • use u64::encode_bytes and u64::decode_bytes for 10-byte encoding & decoding
• data::types: Add more exchanges, and add PERIOD.
• Add the encoder and decoder of bbo and kline.
• file::writer: Add tracing log.
• file::writer: Refactored for extensibility.
• Change to wmjtyd's crypto-crawler and crypto-msg-parser.
• (beta 5) Add methods to serialize and deserialize UNIX timestamp to 6-byte u64.

0.2.0 (beta) - CI

• Also test if cargo doc can generate correct documentation.
• Run cargo clippy and cargo fmt --check

0.2.0 (beta) - Examples

• Add a production example of file::writer.
• (beta 4) data::hex add the testcase of i8 → u8 and u8 → i8

0.2.0 (beta) - Bug fixes

• file::writer: Don't create the directory if it has been existed.
• file::writer: .stop() can't work properly.
• (beta 4) {InfoTypeExpr,ExchangeTypeRepr}::try_from_str should be from lowercase
  • Added the test case of this.
• (beta 5) UNIX timestamp should serialize and deserialize to 6-byte u64.

0.2.0-beta.4

0.2.0 (beta)

0.2.0 (beta) - Breaking changes

• ReceivedTimestamp field is now represented as seconds to fit in u64.
• Symble is now renamed to Symbol.
• Symbol field is now serialized as the big endian representation of u8.
• Map EXCHANGE and INFOTYPE are replaced to the enumeration Exchange and InfoType.
  • Use strum to provide the map-like function.
• Some methods in data::hex are removed.
  • Replace your current code to the native method (.to_be_bytes())
• Removed and added some new unused error variants.

0.2.0 (beta) - Features

• data::orderbook, data::trade: Use BufReader instead of our based-on-Atomic seek reader.
• data::orderbook, data::trade: Use data::fields to serialize & deserialize.
• Separate the serialization & deserialization of fields to data::fields
• Replace the usize converter of hex::encode_num_to_bytes to a robust implementation
  • Related: Some methods in data::hex are removed.
• data::types: Add Exchange and InfoType while removing EXCHANGE and INFOTYPE.
• data::orderbook: Improve the logic of generate_diff.
• data::hex: Deprecate old inextensible methods (encode_num_to_bytes & decode_num_to_bytes)
  • use u32::encode_bytes and u32::decode_bytes for 5-byte encoding & decoding
• data::hex: Add the u64 to 10-byte encode & decode support
  • use u64::encode_bytes and u64::decode_bytes for 10-byte encoding & decoding
• data::types: Add more exchanges, and add PERIOD.
• Add the encoder and decoder of bbo and kline.
• file::writer: Add tracing log.
• file::writer: Refactored for extensibility.
• Change to wmjtyd's crypto-crawler and crypto-msg-parser.

0.2.0 (beta) - CI

• Also test if cargo doc can generate correct documentation.
• Run cargo clippy and cargo fmt --check

0.2.0 (beta) - Examples

• Add a production example of file::writer.
• (beta 4) data::hex add the testcase of i8 → u8 and u8 → i8

0.2.0 (beta) - Bug fixes

• file::writer: Don't create the directory if it has been existed.
• file::writer: .stop() can't work properly.
• (beta 4) {InfoTypeExpr,ExchangeTypeRepr}::try_from_str should be from lowercase
  • Added the test case of this.

0.2.0-beta.2

0.2.0 (beta)

0.2.0 (beta) - Breaking changes

• ReceivedTimestamp field is now represented as seconds to fit in u64.
• Symble is now renamed to Symbol.
• Symbol field is now serialized as the big endian representation of u8.
• Map EXCHANGE and INFOTYPE are replaced to the enumeration Exchange and InfoType.
  • Use strum to provide the map-like function.
• Some methods in data::hex are removed.
  • Replace your current code to the native method (.to_be_bytes())
• Removed and added some new unused error variants.

0.2.0 (beta) - Features

• data::orderbook, data::trade: Use BufReader instead of our based-on-Atomic seek reader.
• data::orderbook, data::trade: Use data::fields to serialize & deserialize.
• Separate the serialization & deserialization of fields to data::fields
• Replace the usize converter of hex::encode_num_to_bytes to a robust implementation
  • Related: Some methods in data::hex are removed.
• data::types: Add Exchange and InfoType while removing EXCHANGE and INFOTYPE.
• data::orderbook: Improve the logic of generate_diff.
• data::hex: Deprecate old inextensible methods (encode_num_to_bytes & decode_num_to_bytes)
  • use u32::encode_bytes and u32::decode_bytes for 5-byte encoding & decoding
• data::hex: Add the u64 to 10-byte encode & decode support
  • use u64::encode_bytes and u64::decode_bytes for 10-byte encoding & decoding
• data::types: Add more exchanges, and add PERIOD.
• Add the encoder and decoder of bbo and kline.
• file::writer: Add tracing log.
• file::writer: Refactored for extensibility.
• Change to wmjtyd's crypto-crawler and crypto-msg-parser.

0.2.0 (beta) - CI

• Also test if cargo doc can generate correct documentation.
• Run cargo clippy and cargo fmt --check

0.2.0 (beta) - Examples

• Add a production example of file::writer.

0.2.0 (beta) - Bug fixes

• file::writer: Don't create the directory if it has been existed.
• file::writer: .stop() can't work properly.

0.2.0-beta.1

0.2.0 (beta)

0.2.0 (beta) - Breaking changes

• ReceivedTimestamp field is now represented as seconds to fit in u64.
• Symble is now renamed to Symbol.
• Symbol field is now serialized as the big endian representation of u8.
• Map EXCHANGE and INFOTYPE are replaced to the enumeration Exchange and InfoType.
  • Use strum to provide the map-like function.
• Some methods in data::hex are removed.
  • Replace your current code to the native method (.to_be_bytes())
• Removed and added some new unused error variants.

0.2.0 (beta) - Features

• data::orderbook, data::trade: Use BufReader instead of our based-on-Atomic seek reader.
• data::orderbook, data::trade: Use data::fields to serialize & deserialize.
• Separate the serialization & deserialization of fields to data::fields
• Replace the usize converter of hex::encode_num_to_bytes to a robust implementation
  • Related: Some methods in data::hex are removed.
• data::types: Add Exchange and InfoType while removing EXCHANGE and INFOTYPE.
• data::orderbook: Improve the logic of generate_diff.
• data::hex: Deprecate old inextensible methods (encode_num_to_bytes & decode_num_to_bytes)
  • use u32::encode_bytes and u32::decode_bytes for 5-byte encoding & decoding
• data::hex: Add the u64 to 10-byte encode & decode support
  • use u64::encode_bytes and u64::decode_bytes for 10-byte encoding & decoding
• data::types: Add more exchanges, and add PERIOD.
• Add the encoder and decoder of bbo and kline.
• file::writer: Add tracing log.
• file::writer: Refactored for extensibility.

0.2.0 (beta) - CI

• Also test if cargo doc can generate correct documentation.
• Run cargo clippy and cargo fmt --check

0.2.0 (beta) - Examples

• Add a production example of file::writer.

0.2.0 (beta) - Bug fixes

• file::writer: Don't create the directory if it has been existed.
• file::writer: .stop() can't work properly.