package for bonial challenge
Project description
flyingtrain - Document
Package that uses an iterative parser to retrieve transport models and the total passenger capacity from a long JSON transport list in a txt file
Installation
This package can be installed with pip
Copy-paste and run this command in the terminal
pip install flyingtrain
Docker
- This project is also dockerized, Docker needs to be installed to run this project in containerization method.
- The Dockerfile uses
python:2
as base image. - There are some feasible commands as indicated in Makefile, or simply execute
make help
, it will show the Make commands that can be used. (We will go through more in detail later)
Tool
This project uses ijson as an iterative JSON parser to avoid dumping the entire data file into memory
Usage
After installation, the following snippet can be used inside a virtual environment which runs the data extraction
import flyingtrain
test_file = 'test.txt' # the full path of the file
flyingtrain.extract_data(test_file)
the result
(flyingtrain) chuhsuan@ubuntu:~/Desktop$ python
Python 2.7.12 (default, Nov 12 2018, 14:36:49)
[GCC 5.4.0 20160609] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import flyingtrain
>>> flyingtrain.extract_data('test.txt')
"planes": 524
"trains": 150
"cars": 14
"distinct-cars": 3
"distinct-planes": 2
"distinct-trains": 1
Docker solution
Edit the test_file
in main.py and execute make run
, the file should be put in the same folder with main.py
. Volume binding can be used to avoid copying the file, but taking docker as a supplementary solution, it's not implemented here.
benchmark
The following command is used in the terminal to show how much time it takes to retrieve the data
python -m timeit -s "import flyingtrain" "flyingtrain.extract_data('test.txt')"
the result
1000 loops, best of 3: 684 usec per loop
which means for executing once, it takes around 684 usec
Docker solution
Edit the test_file
in benchmark.py and execute make runbenchmark
, the file should be put in the same folder with benchmark.py
. Again, volume binding is not implemented here.
the result
[0.6676740646362305, 0.6634271144866943, 0.6310489177703857]
which means measuring execution time with 3 repeats counts and each count with 1000 executions, and for average it takes 663 usec per execution
Possible optimizations
- First, for benchmarking, the build-in module
timeit
is used here. There are also some third party packages can be used such as memory_profiler for monitoring memory consumption of a process as well as line-by-line analysis. - Second, when the record amounts scale up, and the model sets of distinct transports keep increasing, that one can take tons of memory and CPU if we still do it naively by keeping a set of the counts for every model around. There's streaming approximate algorithms for this such as HyperLogLog.
- Last but not least, the format of the datasets. Protocol buffers and recordio, or even Cap'n Proto will be a good try. It's a binary storage format which is faster to parse, and resilient to corruption (recordio files are checksummed, and can skip damaged section without losing the whole file.)
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for flyingtrain-0.1.2-py2-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 966fdd5cd42d7dc69467cb8c0260bb8e5bb1cb94a57f4866f5e4f0931977d331 |
|
MD5 | 670b908b9052af0c31f1958bf86acc49 |
|
BLAKE2b-256 | 2f027da51f56375d1ae8d0159d8a8feaae556d5dd63f666962f82d00f0716985 |