web 机器学习，PHP 连接 python

Question

I'm currently working on a ML project with geo data...

In my web I let the models parameters of the machine learning algorithm to the user, then I send those to an Apache server where PHP gets the parameters... In js I use Ajax to make the POST request.

My ML algorithm is made with Python, right now is working with the library argparse to read the parameters that PHP sends (after of a verification) as command trough the function exec()

I have 2 problems with this:

1. If the ML model takes time to calculate the results, the exec function does't wait for them and returns null after some time, but if it's fast everything is ok. I already have the function set_limit_time(0); in my PHP file.

2. In my local machine it doesn't take so much time to calculate results like on the server, but the server has better hardware, so I don't know what is going on there.

---

• PHP 7.0.15
• Python 2.7
• Server Apache/2.4.18 (Ubuntu 16.04.1 LTS)

---

Also, is there a better way to do this?.

Answer 1

I'd like to suggest another approach.

Whenever you call your algorithm from the command line, you have a bootstrapping time: importing libraries (such as numpy), loading datasets, etc. Then you perform your calculations, return a response and clear the memory. So the next time you need another result, you have to go through all the process again.

I'd suggest embedding that algorithm inside a small Flask application (you already know Python and know how to use PHP for web, it shouldn't be that hard.

Since your python web server will have all the libraries and the datasets already loaded, it would be much faster in answering your questions.

And you can access it from PHP doing an HTTP request with curl (it has timeout.).

I think this would be much easier and scalable.

Just my two cents!

Answer 2

I ended up opening a Python Socket and connecting PHP trough it, now the process it's still slower than my local machine but the timeout is not longer a problem.