文件：icd10cm_tabular_2019.xml
文件大小：8.7M

测试 1

将 XML 全部加在到内存，分别使用 ElementTree.fromstring 和 BeautifulSoup(字符串) 构建 10 次文档树，代码如下：

from bs4 import BeautifulSoup
from xml.etree import ElementTree
import time
from tqdm import trange

xml_path = "./icd10cm_tabular_2019.xml"
xml_content = None
with open(xml_path, "rb") as xml_file:
    xml_content = xml_file.read()

start_time = time.time()
for _ in trange(10):
    ElementTree.fromstring(xml_content)
print("ElementTree", round(time.time() - start_time, 2))

start_time = time.time()
for _ in trange(10):
    BeautifulSoup(xml_content, "lxml-xml")
print("BeautifulSoup lxml", round(time.time() - start_time, 2))

测试结果：

测试 2

不加载 XML 到内存，直接读文件，分别使用 ElementTree.parse 和 BeautifulSoup(文件流) 构建 10 次文档树，代码如下：

from bs4 import BeautifulSoup
from xml.etree import ElementTree
import time
from tqdm import trange

xml_path = "/Users/yangxiao/DATA/2019-ICD-10-CM/icd10cm_tabular_2019.xml"

start_time = time.time()
for _ in trange(10):
    with open(xml_path, "rb") as xml_file:
        ElementTree.parse(xml_file)
print("ElementTree", round(time.time() - start_time, 2))

start_time = time.time()
for _ in trange(10):
    with open(xml_path, "rb") as xml_file:
        BeautifulSoup(xml_file, "lxml-xml")
print("BeautifulSoup lxml", round(time.time() - start_time, 2))

测试结果：

结论：

ElementTree 明显快于 BeautifulSoup lxml-xml ，至少快 10 倍以上。
ElementTree.parse 速度快于 ElementTree.fromstring