Downloads tick-by-tick historical trade data from the Investors Exchange (IEX). Specifically, this tool downloads the archived data feed files which IEX uploads daily on a T+1 basis, and supports parsing these files into tabular format.
Inspired by this Python implementation.
- Extract trade data to DataFrames files using DataFrames.jl
- Download progress bar using ProgressMeter.jl
- TradeReportMessage message type only
- Tested only on version 1.6
- TradeReportMessage message type only
- Untested
The package is still in early shape, but I've included the trade download script this package is built to power to show a usage example. You can run this script like this:
julia --threads <number of CPU cores on your machine> trade_download_script.jl /path/to/save_dirIf you don't care about taking advantage of multi-threading or specifying a custom save directory (the default is ./trade_data), you can just run julia trade_download_script.jl.
In the download script, I avoid downloading more than a handful of raw PCAP data files to disk by running downloads asynchronously with downloaded filenames piped into a limited-sized Channel. As downloads complete, the file paths are consumed by multi-threaded parsing code that reads the TradeReportMessage messages, organizes them into a Julia DataTable, and writes them to disk in parquet format. To take advantage of this parallelization and speed up the parsing of literally every TOPS feed message that IEX has issued since mid 2017, it is recommended you include the --threads flag.
As of v0.1.1, this package only parses the trade report messages in any feed it reads. If you want to read from the DEEP feed or the TOPS v1.5 feed, you'll need to overwrite the default value of the protocol_magic_bytes argument in the read_trade_report_messages function.
Example
import InvestorsExchange as IEX
IEX.read_trade_report_messages("/tmp/20210420_IEXTP1_DEEP1.0.pcap.gz"; protocol_magic_bytes=IEX.DEEP_PROTOCOL_ID_1_0)However, since TOPS and DEEP both contain the trade report messages, there is little reason to use DEEP (which tends to be bigger) to parse the trade report messages. You should expect faster download and parse speeds with TOPS, and thus it's recommended to stick with TOPS.