在python中从csv列表创建一个json树

时间:2012-09-27 07:29:07

标签: python csv recursion tree

我正在尝试从python中的一个简单表构建一个json层次结构。

数据如下所示:

id         parent          name
1          10              test-name-1
2          10              test-name-2
3          5               test-name-3
4          none            test-name-4
5          10              test-name-5
6          none            test-name-6
7          1               test-name-7
8          1               test-name-8
9          8               test-name-9
10         4               test-name-10

我正在寻找这样的输出:

{"$4":{"name":"test-name-4","children":{
      "$10":{"name":"test-name-10","children":{
            "$1":{"name":"test-name-1","children":{
                 "$7":{"name":"test-name-7","children":{}},
                 "$8":{"name":"test-name-8","children":{
                      "$9":{"name":"test-name-9","children":{}}}}}},
            "$2":{"name":"test-name-2","children":{}},
            "$5":{"name":"test-name-5","children":{
                 "$3":{"name":"test-name-3","children":{}}}}}}}},
 "$6":{"name":"test-name-6","children":"test-name-6"}}

我不知道会有多少“叶子”或“根”,或者来自csv的行的顺序是什么。我的问题是,有没有办法可以递归地构建字典/列表一个子节点直到父节点?如何从python中树的“叶子”片段生成分层树?

感谢您的帮助!

3 个答案:

答案 0 :(得分:6)

我有一个基于2个循环的解决方案(1个缓存,1个构建),没有JSON编码器,并且可以提供您所需的输出:

>>> import re
>>> from collections import defaultdict
>>> parents = defaultdict(list)
>>> for i, line in enumerate(file_.split('\n')):
    if i != 0 and line.strip():
        id_, parent, name = re.findall(r'[\d\w-]+', line)
        parents[parent].append((id_, name))


>>> parents
defaultdict(<type 'list'>, {'10': [('1', 'test-name-1'), ('2', 'test-name-2'), ('5', 'test-name-5')], 'none': [('4', 'test-name-4'), ('6', 'test-name-6')], '1': [('7', 'test-name-7'), ('8', 'test-name-8')], '5': [('3', 'test-name-3')], '4': [('10', 'test-name-10')], '8': [('9', 'test-name-9')]})

好的,现在我们有了缓存,递归函数很容易构建我们想要的输出:

>>> def build_tree(d, val):
    return {'$' + id_: {'name': name, 'children': build_tree(d, id_)} for id_, name in d[val]}

我们只需要在之前构建的dict上调用它,值为'none',即树根:

>>> from pprint import pprint
>>> pprint(build_tree(parents, 'none'))
{'$4': {'children': {'$10': {'children': {'$1': {'children': {'$7': {'children': {},
                                                                     'name': 'test-name-7'},
                                                              '$8': {'children': {'$9': {'children': {},
                                                                                         'name': 'test-name-9'}},
                                                                     'name': 'test-name-8'}},
                                                 'name': 'test-name-1'},
                                          '$2': {'children': {},
                                                 'name': 'test-name-2'},
                                          '$5': {'children': {'$3': {'children': {},
                                                                     'name': 'test-name-3'}},
                                                 'name': 'test-name-5'}},
                             'name': 'test-name-10'}},
        'name': 'test-name-4'},
 '$6': {'children': {}, 'name': 'test-name-6'}}
>>> 

答案 1 :(得分:3)

要将所有子节点分配给其父节点,可以对节点列表执行两次传递。第一遍将每个节点添加到UserDict。在第二次传递中,每个节点的父节点都保证在UserDict中,因此节点可以添加到其父节点的children

要序列化为JSON,可以使用JSONEncoder

#!/usr/bin/env python

import sys
import json
import UserDict

class Node(object):
    def __init__(self, nid, parent, name):
        self.nid = nid
        self.parent = parent
        self.children = []
        self.name = name

class NodeDict(UserDict.UserDict):
    def addNodes(self, nodes):
        """ Add every node as a child to its parent by doing two passes."""
        for i in (1, 2):
            for node in nodes:
                self.data[node.nid] = node
                if node.parent in self.data.keys():
                    if node.parent != "none" and
                       node not in self.data[node.parent].children:
                        self.data[node.parent].children.append(node)

class NodeJSONEncoder(json.JSONEncoder):
    def default(self, node):
        if type(node) == Node:
            return {"nid":node.nid, "name":node.name, "children":node.children}
        raise TypeError("{} is not an instance of Node".format(node))

if __name__ == "__main__":
    nodes = []

    with open(sys.argv[1]) as f:
        for row in f.readlines()[1:]:
            nid, parent, name = row.split()
            nodes.append(Node(nid, parent, name))

    nodeDict = NodeDict()
    nodeDict.addNodes(nodes)

    rootNodes = [node for nid, node in nodeDict.items()
                 if node.parent == "none"]
    for rootNode in rootNodes:
        print NodeJSONEncoder().encode(rootNode)

结果:

{"name": "test-name-4", "nid": "4", "children":[
     {"name": "test-name-10", "nid": "10", "children":[
         {"name": "test-name-1", "nid": "1", "children":[
            {"name": "test-name-7", "nid": "7", "children": []},
            {"name": "test-name-8", "nid": "8", "children":[
                {"name": "test-name-9", "nid": "9", "children": []}]}]},
         {"name": "test-name-2", "nid": "2", "children": []},
         {"name": "test-name-5", "nid": "5", "children":[
             {"name": "test-name-3", "nid": "3", "children": []}]}]}]}
{"name": "test-name-6", "nid": "6", "children": []}

答案 2 :(得分:0)

给出的答案在python 3.6中对我不起作用,因为Dict.Dict已被弃用。所以我进行了一些更改以使其工作并通过让用户通过命令行指定child_id,parent_id和子名称的列来稍微概括它。请看下面(我刚刚学习并确信这可以改进,但它适用于我的目的)。

""" Converts a CSV file with Parent/Child Hierarchy to a hierarchical JSON file for front-end processing (javascript/DS)

USAGE: csv2json.py <somefile.csv> a b c  (column nrs of a=child_id, b=parent-id, c=name(of child))

ROOT of hierarchy should contain child_id and parent_id = 'none' or blank.  name must exist """

import sys
import json
import csv
#import UserDict
from collections import UserDict

class Node(object):
    def __init__(self, child_id, parent_id, name):
        self.child_id = child_id
        self.parent_id = parent_id
        self.children = []
        self.name = name

class NodeDict(UserDict):
    def addNodes(self, nodes):
        """ Add every node as a child to its parent_id by doing two passes."""
        for i in (1, 2):
            for node in nodes:
                self.data[node.child_id] = node
                if node.parent_id in self.data.keys():
                    if (node.parent_id != "none" or node.parent_id != "") and node not in self.data[node.parent_id].children:
                        self.data[node.parent_id].children.append(node)

class NodeJSONEncoder(json.JSONEncoder):
    def default(self, node):
        if type(node) == Node:
            return {"name":node.name, "children":node.children}
            raise TypeError("{} is not an instance of Node".format(node))

if __name__ == "__main__":
    nodes = []

    with open(sys.argv[1], 'r') as f:
        reader = csv.reader(f)
        for row in reader:
            if not row[int(sys.argv[4])] : #skip if no name/label exists
                continue
            child_id, parent_id, name = row[int(sys.argv[2])] , row[int(sys.argv[3])] , row[int(sys.argv[4])]

            nodes.append(Node(child_id, parent_id, name))

    nodeDict = NodeDict()
    nodeDict.addNodes(nodes)

    rootNodes = [node for child_id, node in nodeDict.items()
                 if (node.parent_id == "none" or node.parent_id == "")]
    for rootNode in rootNodes:
        print(NodeJSONEncoder().encode(rootNode))