Allow adding additional sys.path entries from the CLI (#1036)

Fixes #1027

Author: chrisrink10

CHANGELOG.md

@@ -5,13 +5,17 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
 ## [Unreleased]
+### Added
+ * Added the `-p`/`--include-path` CLI command to prepend entries to the `sys.path` as an alternative to `PYTHONPATH` (#1027)
+ * Added an empty entry to `sys.path` for all CLI entrypoints (`basilisp run`, `basilisp repl`, etc.) (#1027)
+
 ### Changed
  * The compiler will no longer require `Var` indirection for top-level `do` forms unless those forms specify `^:use-var-indirection` metadata (which currently is only used in the `ns` macro) (#1034)
  * nREPL server no longer sends ANSI color escape sequences in exception messages to clients (#1039)
 
 ### Fixed
  * Fix a bug where the compiler would always generate inline function definitions even if the `inline-functions` compiler option is disabled (#1023)
- * Fix a bug where `defrecord`/`deftype` constructors could not be used in the type's methods. (#1025)
+ * Fix a bug where `defrecord`/`deftype` constructors could not be used in the type's methods (#1025)
  * Fix a bug where `keys` and `vals` would fail for records (#1030)
  * Fix a bug where operations on records created by `defrecord` failed for fields whose Python-safe names were mangled by the Python compiler (#1029)
  * Fix incorrect line numbers for compiler exceptions in nREPL when evaluating forms in loaded files (#1037)

docs/cli.rst

@@ -14,6 +14,28 @@ Basilisp exposes all of it's available configuration options as CLI flags and en
 All Basilisp CLI subcommands which include configuration note the available configuration options when the ``-h`` and ``--help`` flags are given.
 Generally the Basilisp CLI configuration options are simple passthroughs that correspond to :ref:`configuration options for the compiler <compiler_configuration>`.
 
+.. _cli_path_configuration:
+
+``PYTHONPATH`` Configuration
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+Basilisp uses the ``PYTHONPATH`` environment variable and :external:py:data:`sys.path` to determine where to look for Basilisp code when :ref:`requiring namespaces <namespace_requires>`.
+Additional values may be set using the ``-p`` (or ``--include-path``) CLI flags.
+Depending on how Basilisp is invoked from the CLI, an additional entry will automatically be added unless explicitly disabled using ``--include-unsafe-path=false``:
+
+* An empty string (which implies the current working directory) will be prepended to the ``sys.path`` in the following cases:
+
+  * Starting a REPL
+  * Running a string of code directly (using ``run -c``)
+  * Running code directly from ``stdin`` (using ``run -``)
+  * Running a namespace directly (using ``run -n``)
+
+* When running a script directly (as by ``run /path/to/script.lpy``), the parent directory of the script will be prepended to ``sys.path``
+
+.. seealso::
+
+   :ref:`pythonpath_configuration`
+
 .. _start_a_repl_session:
 
 Start a REPL Session

docs/reader.rst

@@ -474,9 +474,10 @@ Custom Data Readers
 
 When Basilisp starts it can load data readers from multiple sources.
 
-It will search in :external:py:attr:`sys.path` for files named ``data_readers.lpy`` or else ``data_readers.cljc``; each which must contain a mapping of qualified symbol tags to qualified symbols of function vars.
+It will search in :external:py:data:`sys.path` for files named ``data_readers.lpy`` or else ``data_readers.cljc``; each which must contain a mapping of qualified symbol tags to qualified symbols of function vars.
 
 .. code-block:: clojure
+
     {my/tag my.namespace/tag-handler}
 
 It will also search for any :external:py:class:`importlib.metadata.EntryPoint` in the group ``basilisp_data_readers`` group.

docs/runtime.rst

@@ -101,6 +101,16 @@ See the documentation for :lpy:fn:`require` for more details.
 
    :lpy:fn:`ns-aliases`, :lpy:fn:`ns-interns`, :lpy:fn:`ns-map`, :lpy:fn:`ns-publics`, :lpy:fn:`ns-refers`, :lpy:fn:`ns-unalias`, :lpy:fn:`ns-unmap`, :lpy:fn:`refer`, :lpy:fn:`require`, :lpy:fn:`use`
 
+.. _pythonpath_configuration:
+
+``PYTHONPATH``, ``sys.path``, and Finding Basilisp Namespaces
+#############################################################
+
+Basilisp uses the ``PYTHONPATH`` environment variable and :external:py:data:`sys.path` to determine where to look for Basilisp code when requiring namespaces.
+This is roughly analogous to the Java classpath in Clojure.
+These values may be set manually, but are more often configured by some project management tool such as Poetry or defined in your Python virtualenv.
+These values may also be set via :ref:`cli` arguments.
+
 .. _vars:
 
 Vars

src/basilisp/cli.py

@@ -2,6 +2,7 @@
 import importlib.metadata
 import io
 import os
+import pathlib
 import sys
 import textwrap
 import types
@@ -78,6 +79,23 @@ def bootstrap_repl(ctx: compiler.CompilerContext, which_ns: str) -> types.Module
     return importlib.import_module(REPL_NS)
 
 
+def init_path(args: argparse.Namespace, unsafe_path: str = "") -> None:
+    """Prepend any import group arguments to `sys.path`, including `unsafe_path` (which
+    defaults to the empty string) if --include-unsafe-path is specified."""
+
+    def prepend_once(path: str) -> None:
+        if path in sys.path:
+            return
+        sys.path.insert(0, path)
+
+    for pth in args.include_path or []:
+        p = pathlib.Path(pth).resolve()
+        prepend_once(str(p))
+
+    if args.include_unsafe_path:
+        prepend_once(unsafe_path)
+
+
 def _to_bool(v: Optional[str]) -> Optional[bool]:
     """Coerce a string argument to a boolean value, if possible."""
     if v is None:
@@ -265,6 +283,39 @@ def _add_debug_arg_group(parser: argparse.ArgumentParser) -> None:
     )
 
 
+def _add_import_arg_group(parser: argparse.ArgumentParser) -> None:
+    group = parser.add_argument_group(
+        "path options",
+        description=(
+            "The path options below can be used to control how Basilisp (and Python) "
+            "find your code."
+        ),
+    )
+    group.add_argument(
+        "--include-unsafe-path",
+        action="store",
+        nargs="?",
+        const=True,
+        default=os.getenv("BASILISP_INCLUDE_UNSAFE_PATH", "true"),
+        type=_to_bool,
+        help=(
+            "if true, automatically prepend a potentially unsafe path to `sys.path`; "
+            "setting `--include-unsafe-path=false` is the Basilisp equivalent to "
+            "setting PYTHONSAFEPATH to a non-empty string for CPython's REPL "
+            "(env: BASILISP_INCLUDE_UNSAFE_PATH; default: true)"
+        ),
+    )
+    group.add_argument(
+        "-p",
+        "--include-path",
+        action="append",
+        help=(
+            "path to prepend to `sys.path`; may be specified more than once to "
+            "include multiple paths (env: PYTHONPATH)"
+        ),
+    )
+
+
 def _add_runtime_arg_group(parser: argparse.ArgumentParser) -> None:
     group = parser.add_argument_group(
         "runtime arguments",
@@ -281,8 +332,8 @@ def _add_runtime_arg_group(parser: argparse.ArgumentParser) -> None:
         const=_to_bool(os.getenv("BASILISP_USE_DATA_READERS_ENTRY_POINT", "true")),
         type=_to_bool,
         help=(
-            "If true, Load data readers from importlib entry points in the "
-            '"basilisp_data_readers" group. (env: '
+            "if true, Load data readers from importlib entry points in the "
+            '"basilisp_data_readers" group (env: '
             "BASILISP_USE_DATA_READERS_ENTRY_POINT; default: true)"
         ),
     )
@@ -386,6 +437,7 @@ def nrepl_server(
     args: argparse.Namespace,
 ) -> None:
     basilisp.init(_compiler_opts(args))
+    init_path(args)
     nrepl_server_mod = importlib.import_module(munge(NREPL_SERVER_NS))
     nrepl_server_mod.start_server__BANG__(
         lmap.map(
@@ -422,6 +474,7 @@ def _add_nrepl_server_subcommand(parser: argparse.ArgumentParser) -> None:
         help='the file path where the server port number is output to, defaults to ".nrepl-port".',
     )
     _add_compiler_arg_group(parser)
+    _add_import_arg_group(parser)
     _add_runtime_arg_group(parser)
     _add_debug_arg_group(parser)
 
@@ -432,6 +485,7 @@ def repl(
 ) -> None:
     opts = _compiler_opts(args)
     basilisp.init(opts)
+    init_path(args)
     ctx = compiler.CompilerContext(filename=REPL_INPUT_FILE_PATH, opts=opts)
     prompter = get_prompter()
     eof = object()
@@ -512,6 +566,7 @@ def _add_repl_subcommand(parser: argparse.ArgumentParser) -> None:
         help="default namespace to use for the REPL",
     )
     _add_compiler_arg_group(parser)
+    _add_import_arg_group(parser)
     _add_runtime_arg_group(parser)
     _add_debug_arg_group(parser)
 
@@ -554,17 +609,21 @@ def run(
             cli_args_var.bind_root(vec.vector(args.args))
 
         if args.code:
+            init_path(args)
             eval_str(target, ctx, ns, eof)
         elif args.load_namespace:
             # Set the requested namespace as the *main-ns*
             main_ns_var = core_ns.find(sym.symbol(runtime.MAIN_NS_VAR_NAME))
             assert main_ns_var is not None
             main_ns_var.bind_root(sym.symbol(target))
 
+            init_path(args)
             importlib.import_module(munge(target))
         elif target == STDIN_FILE_NAME:
+            init_path(args)
             eval_stream(io.TextIOWrapper(sys.stdin.buffer, encoding="utf-8"), ctx, ns)
         else:
+            init_path(args, unsafe_path=str(pathlib.Path(target).resolve().parent))
             eval_file(target, ctx, ns)
 
 
@@ -617,6 +676,7 @@ def _add_run_subcommand(parser: argparse.ArgumentParser) -> None:
         help="command line args made accessible to the script as basilisp.core/*command-line-args*",
     )
     _add_compiler_arg_group(parser)
+    _add_import_arg_group(parser)
     _add_runtime_arg_group(parser)
     _add_debug_arg_group(parser)