如何同时捕获和显示ipython（jupyter笔记本）单元格输出？

Question

如何在捕获细胞输出的jupyter notebook同时显示结果？我不想等待一切完成然后运行variable.show()。

Answer 1

我从内置的实现中改编了以下自定义单元魔术%%capture：

from io import StringIO
import sys

from IPython.core import magic_arguments
from IPython.core.magic import Magics, cell_magic, magics_class
from IPython.utils.capture import CapturedIO


class Tee(StringIO):
    def __init__(self, initial_value='', newline='\n', stream=None):
        self.stream = stream
        super().__init__(initial_value, newline)
    
    def write(self, data):
        if self.stream is not None:
            self.stream.write(data)
        
        super().write(data)


class capture_and_print_output(object):
    stdout = True
    stderr = True
    display = True
    
    def __init__(self, stdout=True, stderr=True, display=True):
        self.stdout = stdout
        self.stderr = stderr
        self.display = display
        self.shell = None
    
    def __enter__(self):
        from IPython.core.getipython import get_ipython
        from IPython.core.displaypub import CapturingDisplayPublisher
        from IPython.core.displayhook import CapturingDisplayHook
        
        self.sys_stdout = sys.stdout
        self.sys_stderr = sys.stderr
        
        if self.display:
            self.shell = get_ipython()
            if self.shell is None:
                self.save_display_pub = None
                self.display = False
        
        stdout = stderr = outputs = None
        if self.stdout:
            stdout = sys.stdout = Tee(stream=sys.stdout)
        if self.stderr:
            stderr = sys.stderr = Tee(stream=sys.stderr)
        if self.display:
            self.save_display_pub = self.shell.display_pub
            self.shell.display_pub = CapturingDisplayPublisher()
            outputs = self.shell.display_pub.outputs
            self.save_display_hook = sys.displayhook
            sys.displayhook = CapturingDisplayHook(shell=self.shell,
                                                   outputs=outputs)
        
        return CapturedIO(stdout, stderr, outputs)
    
    def __exit__(self, exc_type, exc_value, traceback):
        sys.stdout = self.sys_stdout
        sys.stderr = self.sys_stderr
        if self.display and self.shell:
            self.shell.display_pub = self.save_display_pub
            sys.displayhook = self.save_display_hook


@magics_class
class CustomMagics(Magics):
    @magic_arguments.magic_arguments()
    @magic_arguments.argument('output', type=str, default='', nargs='?')
    @magic_arguments.argument('--no-stderr', action='store_true')
    @magic_arguments.argument('--no-stdout', action='store_true')
    @magic_arguments.argument('--no-display', action='store_true')
    @cell_magic
    def tee(self, line, cell):
        args = magic_arguments.parse_argstring(self.tee, line)
        out = not args.no_stdout
        err = not args.no_stderr
        disp = not args.no_display
        with capture_and_print_output(out, err, disp) as io:
            self.shell.run_cell(cell)
        if args.output:
            self.shell.user_ns[args.output] = io

您需要注册此单元魔法，通常在启动配置文件中或将其定义为稍后加载的扩展，例如：

from IPython import get_ipython

get_ipython().register_magics(CustomMagics)

现在您可以按如下方式使用它：

>>> %%tee output
    print(42)
42
>>> output.show()
42

需要注意的一件事：它将像原始%%capture那样捕获丰富的输出，但不会在原始单元格中打印它们。如果这很重要，您可能会找到某种方法来调整上述内容以包含丰富的输出，也许可以通过子类化CapturingDisplayPublisher/Hook。