Learn Julia (14): calling python modules

Julia is a very young language compared to Python or Matlab. Thus its ecosystem is not yet as large as that of Python or Matlab. It shall be a very exciting idea for Julia to be able to import existing Python modules and functions, isn't it?

Today I am glad to have successfully fulfilled some graph database related work in Julia by importing some Python modules. So here I will present step by step what I have done just now.

install pkg PyCall

using pkg
Pkg.add("PyCall")

use PyCall with a correct Python env

I am using Anaconda for Python development. And there are three different Python versions on my machine which are managed by Conda envs. So I need to specify the correct Python path in julia ENV parameter and then re-build PyCall.

ENV["PYTHON"] = "D:\\Anaconda3\\envs\\py36\\python.exe"
Pkg.build("PyCall")
using PyCall

import some modules

pyneo = pyimport("py2neo")
pyjson = pyimport("json")

Data preparation

I faked a data file "data.json" with several french expressions. This file contains three french expression objects and each of them has "id","fr","zh" and "category" properties. The content of this file is as follows:

[
    {
        "id": 1,
        "fr": "Tu me dégoutes!",
        "zh": "我讨厌你！ ",
        "category": "plainte"
    }, {
        "id": 2,
        "fr": "Je veux plus jamais te voir !",
        "zh": "我不愿再见到你！ ",
        "category": "plainte"
    }, {
        "id": 3,
        "fr": "T'es dingue !",
        "zh": "你疯了! ",
        "category": "plainte"
    }
]

datafile = open("./data.json", "r");
ss = read("data.json", String);
print(ss)

Now try to parse the file content as a json object using imported json module.

d = pyjson.loads(ss)
println("d: $d")

It worked thus I can use Pyhton's json module for parsing json-format string (I guess there shall still be issues when the string format is not clean, to be tested in future).

call Python neo4j database and test it

I use Py2neo module to connect to my local Neo4j database. And test it with a simple Cypher command. It works. However, I encounter an error while trying to create a Node in the graph. The type of a Node object has been converted into Dict in Julia.

graph = pyneo.Graph("http://localhost:7474",username="neo4j",password="123456")

res = graph.run("match (n) return n limit 10")
print(collect(res))

n = pyneo.Node("testNode", name="test") 
println(typeof(n))

graph.create(n)  # create nodes in the graph db

Output:

Create nodes using Cypher

Since the Cypher execution works properly, thus I have tried executing cmd create (n:expression) set n.id=$id set n.fr=\"$fr\" set n.zh=\"$zh\" return n for each item of the dict generated from json file. It works and I can see the expressions in the Graph.

for i in d
    id = i["id"]
    fr = i["fr"]
    zh = i["zh"]
    res = graph.run("create (n:expression) set n.id=$id set n.fr=\"$fr\" set n.zh=\"$zh\" return n ")
    println(id, collect(res))
end

Console output:

Neo4j snapshot: