A reactive notebook for Python — run reproducible experiments, query with SQL, execute as a script, deploy as an app, and version with git. Stored as pure Python. All in a modern, AI-native editor.
Python artificial-intelligencedagdata-sciencedata-visualizationdataflowdeveloper-toolsmachine-learningnotebookspipelinepythonreactivesqlweb-app
fix: no diff churn when adding requires-python to a sandboxed file (#8054)
Akshay Agrawal419d77a545ba
diff --git a/marimo/_cli/sandbox.py b/marimo/_cli/sandbox.pyindex af3b8a54e9..6e1363927f 100644--- a/marimo/_cli/sandbox.py+++ b/marimo/_cli/sandbox.py@@ -24,7 +24,6 @@ from marimo._utils.inline_script_metadata import (has_marimo_in_script_metadata,is_marimo_dependency,)-from marimo._utils.scripts import with_python_version_requirementfrom marimo._utils.uv import find_uv_binfrom marimo._utils.versions import is_editablefrom marimo._version import __version__@@ -324,14 +323,12 @@ def construct_uv_command(def _ensure_python_version_in_script_metadata(name: str) -> None:"""Add requires-python to script metadata if not present.- Reads the file, parses the PEP 723 metadata, adds requires-python- if missing, and writes the updated metadata back.+ Inserts a requires-python line directly into the existing PEP 723+ metadata block without re-serializing, to avoid reformatting diffs."""- from marimo._utils.scripts import (- REGEX,- read_pyproject_from_script,- write_pyproject_to_script,- )+ import re++ from marimo._utils.scripts import read_pyproject_from_scriptwith open(name, encoding="utf-8") as f:content = f.read()@@ -342,17 +339,22 @@ def _ensure_python_version_in_script_metadata(name: str) -> None:returnif "requires-python" in project:- # Already has Python versionreturn- # Generate new script metadata block- project = with_python_version_requirement(project)- new_block = write_pyproject_to_script(project)-- # Replace the old block with the new one- import re+ version_tuple = platform.python_version_tuple()+ requires_line = (+ f'# requires-python = ">={version_tuple[0]}.{version_tuple[1]}"\n'+ )- new_content = re.sub(REGEX, new_block, content, count=1)+ # Insert directly after the opening "# /// script" marker to avoid+ # re-serializing the entire block and causing formatting churn.+ new_content = re.sub(+ r"^# /// script$",+ "# /// script\n" + requires_line.rstrip(),+ content,+ count=1,+ flags=re.MULTILINE,+ )if new_content != content:with open(name, "w", encoding="utf-8") as f: