VS Code Flask 调试运行比从命令行运行 Flask 慢得多（具有相同的参数）

Question

I am currently in early developing stages of a flask app in VS Code and I don't understand why I have huge runtime differences between starting the flask app using the VS Code debugger and when I start my flask app from the command line(5 to 6 times slower in VS code debugging), but with the same flask parameters. This seems to mainly happen when I use pandas/numpy objects.

I have the following content in my launch.json (only content relevant to "Python: Flask"

        {
            "name": "Python: Flask",
            "type": "python",
            "request": "launch",
            "module": "flask",
            "env": {
                "FLASK_APP": "test_flask_app.py",
                "FLASK_ENV": "development",
            },
            "args": [
                "run",
                "--no-debugger",
                "--no-reload",
            ],
            "jinja": true
        },

I then click on "Start debugging" and the following is displayed in the terminal:

 * Serving Flask app "test_flask_app.py"
 * Environment: development
 * Debug mode: on
 * Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)

When I start Flask from the command line with the following commands: export FLASK_APP=test_flask_app.py export FLASK_ENV=devlopment flask run --no-debugger --no-reload

The following information is displayed on startup (identical to the first one):

 * Serving Flask app "test_flask_app.py"
 * Environment: development
 * Debug mode: on
 * Running on http://127.0.0.1:5000/ (Press CTRL+C to quit)

The strange thing is that the runtime time of one page takes about 800ms when I run Flask directly from the terminal and 5 to 6 seconds when I run Flask from VS Code debugger.

I believe all my parameters are identical (and I can of course start my Flask manually), but I would like to understand why there is such a big difference in runtimes?

Here is the sample code I used:

import time
import numpy as np
import pandas as pd
from flask import Flask

# vectorized function
def generateHeaderObject(text, level, calc, axis):
    return "." * int(level) + text + '|' + str(False) + '|' + "" + '|' + \
        "hdr " + axis + "ht" + (" " + axis + "agg" if calc else "")

# create sample dataframe
n = 1000000
df = pd.DataFrame(dict(
    calc=np.random.choice(a=[False,True],size=n),
    level=np.random.randint(1, 12, size=n),
    id=np.random.randint(10000000, 99999999, size=n)
))
df['id'] = df['id'].astype(str)

app = Flask(__name__) 

@app.route('/') 
# ‘/’ URL is bound with hello_world() function. 
def main():
    # Calculate new column
    start = time.time()
    df['output'] = np.vectorize(generateHeaderObject)(
        df['id'], df['level'], df['calc'], "r"
    )
    return 'runtime np.vectorize (ms): ' + str((time.time() - start) * 1000)

Answer 1

This is simply overhead from using the debugger due to it having to run within your Flask application. If you use "Run without Debugging" in VS Code you won't see the slowdown.