Make `pydabs/venv_path` optional (#1687)

## Changes
Make `pydabs/venv_path` optional. When not specified, CLI detects the
Python interpreter using `python.DetectExecutable`, the same way as for
`artifacts`. `python.DetectExecutable` works correctly if a virtual
environment is activated or `python3` is available on PATH through other
means.

Extract the venv detection code from PyDABs into `libs/python/detect`.
This code will be used when we implement the `python/venv_path` section
in `databricks.yml`.

## Tests
Unit tests and manually

---------

Co-authored-by: Pieter Noordhuis <pcnoordhuis@gmail.com>

bundle/artifacts/whl/infer.go

@@ -15,6 +15,8 @@ type infer struct {
 
 func (m *infer) Apply(ctx context.Context, b *bundle.Bundle) diag.Diagnostics {
 	artifact := b.Config.Artifacts[m.name]
+
+	// TODO use python.DetectVEnvExecutable once bundle has a way to specify venv path
 	py, err := python.DetectExecutable(ctx)
 	if err != nil {
 		return diag.FromErr(err)

bundle/config/experimental.go

@@ -36,8 +36,8 @@ type PyDABs struct {
 
 	// VEnvPath is path to the virtual environment.
 	//
-	// Required if PyDABs is enabled. PyDABs will load the code in the specified
-	// environment.
+	// If enabled, PyDABs will execute code within this environment. If disabled,
+	// it defaults to using the Python interpreter available in the current shell.
 	VEnvPath string `json:"venv_path,omitempty"`
 
 	// Import contains a list Python packages with PyDABs code.

bundle/config/mutator/python/python_mutator.go

@@ -7,8 +7,8 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
-	"runtime"
 
+	"github.com/databricks/cli/libs/python"
 	"github.com/databricks/databricks-sdk-go/logger"
 
 	"github.com/databricks/cli/bundle/env"
@@ -86,23 +86,15 @@ func (m *pythonMutator) Apply(ctx context.Context, b *bundle.Bundle) diag.Diagno
 		return nil
 	}
 
-	if experimental.PyDABs.VEnvPath == "" {
-		return diag.Errorf("\"experimental.pydabs.enabled\" can only be used when \"experimental.pydabs.venv_path\" is set")
-	}
-
 	// mutateDiags is used because Mutate returns 'error' instead of 'diag.Diagnostics'
 	var mutateDiags diag.Diagnostics
 	var mutateDiagsHasError = errors.New("unexpected error")
 
 	err := b.Config.Mutate(func(leftRoot dyn.Value) (dyn.Value, error) {
-		pythonPath := interpreterPath(experimental.PyDABs.VEnvPath)
+		pythonPath, err := detectExecutable(ctx, experimental.PyDABs.VEnvPath)
 
-		if _, err := os.Stat(pythonPath); err != nil {
-			if os.IsNotExist(err) {
-				return dyn.InvalidValue, fmt.Errorf("can't find %q, check if venv is created", pythonPath)
-			} else {
-				return dyn.InvalidValue, fmt.Errorf("can't find %q: %w", pythonPath, err)
-			}
+		if err != nil {
+			return dyn.InvalidValue, fmt.Errorf("failed to get Python interpreter path: %w", err)
 		}
 
 		cacheDir, err := createCacheDir(ctx)
@@ -423,11 +415,16 @@ func isOmitemptyDelete(left dyn.Value) bool {
 	}
 }
 
-// interpreterPath returns platform-specific path to Python interpreter in the virtual environment.
-func interpreterPath(venvPath string) string {
-	if runtime.GOOS == "windows" {
-		return filepath.Join(venvPath, "Scripts", "python3.exe")
-	} else {
-		return filepath.Join(venvPath, "bin", "python3")
+// detectExecutable lookups Python interpreter in virtual environment, or if not set, in PATH.
+func detectExecutable(ctx context.Context, venvPath string) (string, error) {
+	if venvPath == "" {
+		interpreter, err := python.DetectExecutable(ctx)
+		if err != nil {
+			return "", err
+		}
+
+		return interpreter, nil
 	}
+
+	return python.DetectVEnvExecutable(venvPath)
 }

bundle/config/mutator/python/python_mutator_test.go

@@ -282,7 +282,7 @@ func TestPythonMutator_venvRequired(t *testing.T) {
 }
 
 func TestPythonMutator_venvNotFound(t *testing.T) {
-	expectedError := fmt.Sprintf("can't find %q, check if venv is created", interpreterPath("bad_path"))
+	expectedError := fmt.Sprintf("failed to get Python interpreter path: can't find %q, check if virtualenv is created", interpreterPath("bad_path"))
 
 	b := loadYaml("databricks.yml", `
       experimental:
@@ -596,9 +596,7 @@ func loadYaml(name string, content string) *bundle.Bundle {
 	}
 }
 
-func withFakeVEnv(t *testing.T, path string) {
-	interpreterPath := interpreterPath(path)
-
+func withFakeVEnv(t *testing.T, venvPath string) {
 	cwd, err := os.Getwd()
 	if err != nil {
 		panic(err)
@@ -608,6 +606,8 @@ func withFakeVEnv(t *testing.T, path string) {
 		panic(err)
 	}
 
+	interpreterPath := interpreterPath(venvPath)
+
 	err = os.MkdirAll(filepath.Dir(interpreterPath), 0755)
 	if err != nil {
 		panic(err)
@@ -618,9 +618,22 @@ func withFakeVEnv(t *testing.T, path string) {
 		panic(err)
 	}
 
+	err = os.WriteFile(filepath.Join(venvPath, "pyvenv.cfg"), []byte(""), 0755)
+	if err != nil {
+		panic(err)
+	}
+
 	t.Cleanup(func() {
 		if err := os.Chdir(cwd); err != nil {
 			panic(err)
 		}
 	})
 }
+
+func interpreterPath(venvPath string) string {
+	if runtime.GOOS == "windows" {
+		return filepath.Join(venvPath, "Scripts", "python3.exe")
+	} else {
+		return filepath.Join(venvPath, "bin", "python3")
+	}
+}

libs/python/detect.go

@@ -3,9 +3,23 @@ package python
 import (
 	"context"
 	"errors"
+	"fmt"
+	"io/fs"
+	"os"
 	"os/exec"
+	"path/filepath"
+	"runtime"
 )
 
+// DetectExecutable looks up the path to the python3 executable from the PATH
+// environment variable.
+//
+// If virtualenv is activated, executable from the virtualenv is returned,
+// because activating virtualenv adds python3 executable on a PATH.
+//
+// If python3 executable is not found on the PATH, the interpreter with the
+// least version that satisfies minimal 3.8 version is returned, e.g.
+// python3.10.
 func DetectExecutable(ctx context.Context) (string, error) {
 	// TODO: add a shortcut if .python-version file is detected somewhere in
 	// the parent directory tree.
@@ -32,3 +46,35 @@ func DetectExecutable(ctx context.Context) (string, error) {
 	}
 	return interpreter.Path, nil
 }
+
+// DetectVEnvExecutable returns the path to the python3 executable inside venvPath,
+// that is not necessarily activated.
+//
+// If virtualenv is not created, or executable doesn't exist, the error is returned.
+func DetectVEnvExecutable(venvPath string) (string, error) {
+	interpreterPath := filepath.Join(venvPath, "bin", "python3")
+	if runtime.GOOS == "windows" {
+		interpreterPath = filepath.Join(venvPath, "Scripts", "python3.exe")
+	}
+
+	if _, err := os.Stat(interpreterPath); err != nil {
+		if errors.Is(err, fs.ErrNotExist) {
+			return "", fmt.Errorf("can't find %q, check if virtualenv is created", interpreterPath)
+		} else {
+			return "", fmt.Errorf("can't find %q: %w", interpreterPath, err)
+		}
+	}
+
+	// pyvenv.cfg must be always present in correctly configured virtualenv,
+	// read more in https://snarky.ca/how-virtual-environments-work/
+	pyvenvPath := filepath.Join(venvPath, "pyvenv.cfg")
+	if _, err := os.Stat(pyvenvPath); err != nil {
+		if errors.Is(err, fs.ErrNotExist) {
+			return "", fmt.Errorf("expected %q to be virtualenv, but pyvenv.cfg is missing", venvPath)
+		} else {
+			return "", fmt.Errorf("can't find %q: %w", pyvenvPath, err)
+		}
+	}
+
+	return interpreterPath, nil
+}

libs/python/detect_test.go

@@ -0,0 +1,46 @@
+package python
+
+import (
+	"os"
+	"path/filepath"
+	"runtime"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestDetectVEnvExecutable(t *testing.T) {
+	dir := t.TempDir()
+	interpreterPath := interpreterPath(dir)
+
+	err := os.Mkdir(filepath.Dir(interpreterPath), 0755)
+	require.NoError(t, err)
+
+	err = os.WriteFile(interpreterPath, []byte(""), 0755)
+	require.NoError(t, err)
+
+	err = os.WriteFile(filepath.Join(dir, "pyvenv.cfg"), []byte(""), 0755)
+	require.NoError(t, err)
+
+	executable, err := DetectVEnvExecutable(dir)
+
+	assert.NoError(t, err)
+	assert.Equal(t, interpreterPath, executable)
+}
+
+func TestDetectVEnvExecutable_badLayout(t *testing.T) {
+	dir := t.TempDir()
+
+	_, err := DetectVEnvExecutable(dir)
+
+	assert.Errorf(t, err, "can't find %q, check if virtualenv is created", interpreterPath(dir))
+}
+
+func interpreterPath(venvPath string) string {
+	if runtime.GOOS == "windows" {
+		return filepath.Join(venvPath, "Scripts", "python3.exe")
+	} else {
+		return filepath.Join(venvPath, "bin", "python3")
+	}
+}