"""Utilities for using Manim with IPython (in particular: Jupyter notebooks)"""
from __future__ import annotations
import mimetypes
import shutil
from datetime import datetime
from pathlib import Path
from typing import Any
from manim import config, logger, tempconfig
from manim.__main__ import main
from manim.renderer.shader import shader_program_cache
from ..constants import RendererType
__all__ = ["ManimMagic"]
try:
from IPython import get_ipython
from IPython.core.interactiveshell import InteractiveShell
from IPython.core.magic import (
Magics,
line_cell_magic,
magics_class,
needs_local_scope,
)
from IPython.display import Image, Video, display
except ImportError:
pass
else:
@magics_class
class ManimMagic(Magics):
def __init__(self, shell: InteractiveShell) -> None:
super().__init__(shell)
self.rendered_files: dict[Path, Path] = {}
@needs_local_scope
@line_cell_magic
def manim(
self,
line: str,
cell: str | None = None,
local_ns: dict[str, Any] | None = None,
) -> None:
r"""Render Manim scenes contained in IPython cells.
Works as a line or cell magic.
.. hint::
This line and cell magic works best when used in a JupyterLab
environment: while all of the functionality is available for
classic Jupyter notebooks as well, it is possible that videos
sometimes don't update on repeated execution of the same cell
if the scene name stays the same.
This problem does not occur when using JupyterLab.
Please refer to ``_ for more information about JupyterLab
and Jupyter notebooks.
Usage in line mode::
%manim [CLI options] MyAwesomeScene
Usage in cell mode::
%%manim [CLI options] MyAwesomeScene
class MyAweseomeScene(Scene):
def construct(self):
...
Run ``%manim --help`` and ``%manim render --help`` for possible command line interface options.
.. note::
The maximal width of the rendered videos that are displayed in the notebook can be
configured via the ``media_width`` configuration option. The default is set to ``25vw``,
which is 25% of your current viewport width. To allow the output to become as large
as possible, set ``config.media_width = "100%"``.
The ``media_embed`` option will embed the image/video output in the notebook. This is
generally undesirable as it makes the notebooks very large, but is required on some
platforms (notably Google's CoLab, where it is automatically enabled unless suppressed
by ``config.embed = False``) and needed in cases when the notebook (or converted HTML
file) will be moved relative to the video locations. Use-cases include building
documentation with Sphinx and JupyterBook. See also the :mod:`manim directive for Sphinx
`.
Examples
--------
First make sure to put ``import manim``, or even ``from manim import *``
in a cell and evaluate it. Then, a typical Jupyter notebook cell for Manim
could look as follows::
%%manim -v WARNING --disable_caching -qm BannerExample
config.media_width = "75%"
config.media_embed = True
class BannerExample(Scene):
def construct(self):
self.camera.background_color = "#ece6e2"
banner_large = ManimBanner(dark_theme=False).scale(0.7)
self.play(banner_large.create())
self.play(banner_large.expand())
Evaluating this cell will render and display the ``BannerExample`` scene defined in the body of the cell.
.. note::
In case you want to hide the red box containing the output progress bar, the ``progress_bar`` config
option should be set to ``None``. This can also be done by passing ``--progress_bar None`` as a
CLI flag.
"""
if cell:
exec(cell, local_ns)
args = line.split()
if not len(args) or "-h" in args or "--help" in args or "--version" in args:
main(args, standalone_mode=False, prog_name="manim")
return
modified_args = self.add_additional_args(args)
args = main(modified_args, standalone_mode=False, prog_name="manim")
assert isinstance(local_ns, dict)
with tempconfig(local_ns.get("config", {})):
config.digest_args(args)
renderer = None
if config.renderer == RendererType.OPENGL:
from manim.renderer.opengl_renderer import OpenGLRenderer
renderer = OpenGLRenderer()
try:
SceneClass = local_ns[config["scene_names"][0]]
scene = SceneClass(renderer=renderer)
scene.render()
finally:
# Shader cache becomes invalid as the context is destroyed
shader_program_cache.clear()
# Close OpenGL window here instead of waiting for the main thread to
# finish causing the window to stay open and freeze
if renderer is not None and renderer.window is not None:
renderer.window.close()
if config["output_file"] is None:
logger.info("No output file produced")
return
local_path = Path(config["output_file"]).relative_to(Path.cwd())
tmpfile = (
Path(config["media_dir"])
/ "jupyter"
/ f"{_generate_file_name()}{local_path.suffix}"
)
if local_path in self.rendered_files:
self.rendered_files[local_path].unlink()
self.rendered_files[local_path] = tmpfile
tmpfile.parent.mkdir(parents=True, exist_ok=True)
shutil.copy(local_path, tmpfile)
file_type = mimetypes.guess_type(config["output_file"])[0]
assert isinstance(file_type, str)
embed = config["media_embed"]
if not embed:
# videos need to be embedded when running in google colab.
# do this automatically in case config.media_embed has not been
# set explicitly.
embed = "google.colab" in str(get_ipython())
if file_type.startswith("image"):
result = Image(filename=config["output_file"])
else:
result = Video(
tmpfile,
html_attributes=f'controls autoplay loop style="max-width: {config["media_width"]};"',
embed=embed,
)
display(result)
def add_additional_args(self, args: list[str]) -> list[str]:
additional_args = ["--jupyter"]
# Use webm to support transparency
if "-t" in args and "--format" not in args:
additional_args += ["--format", "webm"]
return additional_args + args[:-1] + [""] + [args[-1]]
def _generate_file_name() -> str:
val: str = (
config["scene_names"][0] + "@" + datetime.now().strftime("%Y-%m-%d@%H-%M-%S")
)
return val