Merge branch 'main' into precompile-first-call

Author: cjdoris

.github/workflows/register-pypi.yml

@@ -17,7 +17,7 @@ jobs:
     - name: Set up Python
       uses: actions/setup-python@v7
       with:
-        python-version: "3.10"
+        python-version: "3.11"
     - name: Install build
       run: >-
         python -m

.github/workflows/tests.yml

@@ -41,6 +41,17 @@ jobs:
 
       - uses: julia-actions/cache@v3
 
+      # The package cache includes Conda's build configuration, but not its root environment.
+      - name: Rebuild Conda configuration
+        shell: julia --color=yes {0}
+        env:
+          JULIA_PKG_PRECOMPILE_AUTO: '0'
+        run: |
+          using Pkg
+          Pkg.activate(; temp=true)
+          Pkg.add("Conda")
+          Pkg.build("Conda")
+
       - name: Build package
         uses: julia-actions/julia-buildpkg@v1
         env:
@@ -83,7 +94,7 @@ jobs:
       fail-fast: false
       matrix:
         os: [ubuntu-latest, windows-latest, macos-latest]
-        pyversion: ['3', '3.10']
+        pyversion: ['3', '3.11']
         juliaexe: ["@JuliaPkg"]
         include:
         - os: ubuntu-latest

AGENTS.md

@@ -9,5 +9,10 @@
 
 The majority of tests live in the Julia package; Python tests cover functionality that cannot be exercised from Julia (e.g., JuliaCall-specific behavior). Run both suites—typically Julia first—in whichever order makes sense.
 
+- In restricted sandboxes, both suites need approval to write to the shared caches (`~/.julia` and `~/.cache/uv`).
+- `pysrc/juliacall/juliapkg.json` is tracked; restore its original contents after using the development copy for Python tests.
+
 ## Meta instructions
+- When testing an older Julia version, a local manifest generated by newer Julia can fail on stdlib dependencies. Use a temporary environment with `Pkg.develop(path=pwd()); Pkg.test("PythonCall")` to resolve dependencies for that Julia version.
+- CI caches Conda.jl build configuration without its root environment. Rebuild Conda after cache restore before PyCall is built; otherwise Julia 1.10 can fail before tests with "Path to conda environment is not valid".
 - When you discover environment quirks, false assumptions, process fixes, or any other generally useful info, update this AGENTS.md so future coding agents have the information.

CHANGELOG.md

@@ -1,8 +1,11 @@
 # Changelog
 
 ## Unreleased
-* The first call from Python in a new JuliaCall session no longer compiles
-  `jl.<name>` lookup on a module or `jl.seval`, saving about a second.
+* Improved TTFX in JuliaCall.
+* Minimum supported Python version is now 3.11.
+
+## 0.9.37 (2026-10-12)
+* Compatibility fix with Python 3.15 from conda-forge, which no longer bundles libpython.
 * Bug fixes.
 
 ## 0.9.36 (2026-09-18)

CondaPkg.toml

@@ -1,18 +1,19 @@
-[deps]
-conda = "=26.7.2"
+[deps.openssl]
+version = "<=julia"
 
-    [deps.openssl]
-    version = "<=julia"
+[deps.libstdcxx]
+version = "<=julia"
 
-    [deps.libstdcxx]
-    version = "<=julia"
+[deps.libstdcxx-ng]
+version = "<=julia"
 
-    [deps.libstdcxx-ng]
-    version = "<=julia"
+[deps.python]
+build = "**cpython**"
+version = ">=3.11,!=3.14.0,!=3.14.1,<4"
 
-    [deps.python]
-    build = "**cpython**"
-    version = ">=3.10,!=3.14.0,!=3.14.1,<4"
+[deps.libpython]
+build = "**bundled-or-separate**"
+channel = "conda-forge"
 
 [dev.deps]
 numpy = ""

Project.toml

@@ -1,7 +1,7 @@
 name = "PythonCall"
 uuid = "6099a3de-0909-46bc-b1f4-468b9a2dfc0d"
 authors = ["Christopher Doris <github.com/cjdoris>"]
-version = "0.9.36"
+version = "0.9.37"
 
 [deps]
 CondaPkg = "992eb4ea-22a4-4c89-a5bb-47a3300528ab"
@@ -16,7 +16,7 @@ UnsafePointers = "e17b2a0c-0bdf-430a-bd0c-3a23cae4ff39"
 
 [compat]
 CategoricalArrays = "0.10, 1"
-CondaPkg = "0.2.33"
+CondaPkg = "0.2.38"
 Dates = "1"
 Libdl = "1"
 MacroTools = "0.5"

README.md

@@ -14,7 +14,7 @@ Bringing [**Python®**](https://www.python.org/) and [**Julia**](https://juliala
 - Fast non-copying conversion of numeric arrays in either direction: modify Python arrays (e.g. `bytes`, `array.array`, `numpy.ndarray`) from Julia or Julia arrays from Python.
 - Helpful wrappers: interpret Python sequences, dictionaries, arrays, dataframes and IO streams as their Julia counterparts, and vice versa.
 - Beautiful stack-traces.
-- Supports modern systems: tested on Windows, MacOS and Linux; 64-bit; Julia 1.10 upwards and Python 3.10 upwards.
+- Supports modern systems: tested on Windows, MacOS and Linux; 64-bit; Julia 1.10 upwards and Python 3.11 upwards.
 
 ⭐ If you like this, a GitHub star would be lovely thank you. ⭐
 

benchmark/Project.toml

@@ -1,2 +1,3 @@
 [deps]
 BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
+PythonCall = "6099a3de-0909-46bc-b1f4-468b9a2dfc0d"

docs/src/faq.md

@@ -112,7 +112,7 @@ np = pyimport("numpy")
 ## What versions of Python and Julia do you support?
 
 Each release of PythonCall and JuliaCall will support and require:
-- Any currently supported version of Python, [see here](https://devguide.python.org/versions/). Currently 3.10+.
+- Any currently supported version of Python, [see here](https://devguide.python.org/versions/). Currently 3.11+.
 - The current Julia LTS version and newer, [see here](https://julialang.org/downloads/#long_term_support_release). Currently 1.10+.
 
 Only the latest patch release within each minor version is supported.

docs/src/index.md

@@ -7,4 +7,4 @@ Bringing [**Python®**](https://www.python.org/) and [**Julia**](https://juliala
 - Fast non-copying conversion of numeric arrays in either direction: modify Python arrays (e.g. `bytes`, `array.array`, `numpy.ndarray`) from Julia or Julia arrays from Python.
 - Helpful wrappers: interpret Python sequences, dictionaries, arrays, dataframes and IO streams as their Julia counterparts, and vice versa.
 - Beautiful stack-traces.
-- Works anywhere: tested on Windows, MacOS and Linux; 64-bit; Julia 1.10 upwards and Python 3.10 upwards.
+- Works anywhere: tested on Windows, MacOS and Linux; 64-bit; Julia 1.10 upwards and Python 3.11 upwards.