Python，VIM：卡在<<这里的python对象的名称空间，范围，寿命

Question

If i have two functions:

function! foo()
python3 << HERE
 import mylib
 pass
HERE

function! bar()
python3 << HERE
 import mylib
 pass
HERE

The src says:

 1. Python interpreter main program
 3. Implementation of the Vim module for Python

So, is the python interpreter embedded in vim AND additionally, are separate extensions to python provided (wrapper functions for the VIM API)

How many times is mylib compiled to bytecode and loaded? Does each vimscript function get its own mylib - can I instantiate something and expect it to be visible in the other function? I have a bunch of leader functions that share similar code and act on the same buffer so I wanted to know if I could reuse that data-structure. How many times is the interpreter loaded into memory: once obviously at vim runtime.

Answer 1

1. Python interpreter is indeed loaded once.
2. Python interpreter caches imported modules in sys.modules , so mylib is loaded once per Vim instance.
3. Python interpreter caches compiled bytecode in the *.pyc files on the filesystem, so mylib is compiled once per its update (assuming Python was able to detect this update; usually it has no problems with this).

---

In any case, do never use python << EOF , all plugins should always use namespaced functions: eg

Bad:

python << EOF
from y import foo, bar

def my_function1():
    foo()

def my_function2(i):
    bar(i)

for i in range(2):
    my_function1()
    my_function2(i)
EOF

Good:

# pythonx/mymodule.py
from y import foo, bar

def my_function1():
    foo()

def my_function2(i):
    bar(i)

def start():
    for i in range(2):
        my_function1()
        my_function2(i)

" plugin/myscript.vim
python import mymodule; mymodule.start()

Reasoning:

1. In the above example all of my_function1 , my_function2 , foo , bar , i appear in the __main__ module global namespace. If you install some plugin which does not follow good practices (there are lots of them) and it appears to define different foo or bar your plugin will not work. Or that plugin will not work if it happened to be loaded before yours.
2. python << EOF is recompiled each time it is called.