Implement PythonTA LSP

.vscodeignore

@@ -22,4 +22,5 @@ noxfile.py
 .pylintrc
 **/requirements.txt
 **/requirements.in
-**/tool/_debug_server.py
+**/tool/_debug_server.py
+media/**

CHANGELOG.md

@@ -8,6 +8,7 @@
 
 ### ✨ New features and improvements
 
+- Implement initial PythonTA LSP server
 - Add initial PythonTA implementation that simply calls the CLI
 
 ### 🐛 Bug fixes

README.md

@@ -14,10 +14,36 @@ This project is an extension for running [PythonTA](https://www.cs.toronto.edu/~
 
 ### Setup
 
-1. Install Python dependencies: `uv sync`.
-2. Install Javascript dependencies: `pnpm install`.
+1. Install the extension's dependencies into the bundled libraries directory:
+```bash
+uv pip install --target bundled/libs -r pyproject.toml
+```
+2. Install Python dependencies: `uv sync`.
+3. Install Javascript dependencies: `pnpm install`.
 
-To start the extension, use the `Debug Extension and Python` configuration in VS Code.
+### Testing the Extension Locally
+
+1. To start the extension, use the `Debug Extension and Python` (shortcut `F5`) configuration in VS Code.
+2. In the new VS Code window that appears, open a folder containing Python files, or create a new file.
+3. Open a file with a snippet or write a snippet of Python code that violates standard PythonTA rule. For example:
+   ```python
+   def add_numbers(a, b):
+       return a + c
+    ```
+4. Save the file, and look for the diagnostic messages and rule codes generated by PythonTA.
+5. You can view the extension output under the Output tab, and select "PythonTA VS Code Extension"
+
+### Troubleshooting: Server Crashes on File Reload
+
+If you are actively testing the extension using the debugger (`F5`), you may notice that closing and reopening a Python file causes the server to crash repeatedly with a `ConnectionRefusedError`. 
+
+**This is expected behavior in the development environment and will not happen in the published extension.** It occurs because the VS Code Extension Development Host wraps the language server in a `debugpy` instance that fails to cleanly release its port when the file reloads.
+
+If you want to avoid this during testing, and you do not need breakpoints, launch the Extension Development Host using **Debug Extension Only** (via the dropdown in the Run and Debug menu) instead of `F5`.
+
+### Demo Video
+
+<video controls src="media/demo.mp4" title="Title"></video>
 
 ### Running tests
 

bundled/tool/_debug_server.py

@@ -30,7 +30,7 @@ def update_sys_path(path_to_add: str) -> None:
     # This will ensure that execution is paused as soon as the debugger
     # connects to VS Code. If you don't want to pause here comment this
     # line and set breakpoints as appropriate.
-    debugpy.breakpoint()
+    # debugpy.breakpoint()
 
 SERVER_PATH = os.fspath(pathlib.Path(__file__).parent / "lsp_server.py")
 # NOTE: Set breakpoint in `lsp_server.py` before continuing.

bundled/tool/lsp_server.py

@@ -37,6 +37,7 @@ def update_sys_path(path_to_add: str, strategy: str) -> None:
 # pylint: disable=wrong-import-position,import-error
 import lsp_jsonrpc as jsonrpc
 import lsp_utils as utils
+from lsprotocol import converters
 from lsprotocol import types as lsp
 from pygls import uris, workspace
 from pygls.lsp.server import LanguageServer
@@ -80,12 +81,10 @@ def update_sys_path(path_to_add: str, strategy: str) -> None:
 #  Black: https://github.com/microsoft/vscode-black-formatter/blob/main/bundled/tool
 #  isort: https://github.com/microsoft/vscode-isort/blob/main/bundled/tool
 
-TOOL_MODULE = "python-ta"
+TOOL_MODULE = "python_ta"
 
 TOOL_DISPLAY = "PythonTA"
-
-TOOL_ARGS = []  # default arguments always passed to your tool.
-
+TOOL_ARGS = ["--output-format", "pyta-lsp", "--exit-zero"]  # default arguments always passed to your tool
 
 # TODO: If your tool is a linter then update this section.
 # Delete "Linting features" section if your tool is NOT a linter.
@@ -239,140 +238,47 @@ def _linting_helper(document: workspace.Document) -> list[lsp.Diagnostic]:
     # support linting over stdin to be effective. Read, and update
     # _run_tool_on_document and _run_tool functions as needed for your project.
     result = _run_tool_on_document(document)
-    return _parse_output_using_regex(result.stdout) if result.stdout else []
-
-
-# TODO: If your linter outputs in a known format like JSON, then parse
-# accordingly. But incase you need to parse the output using RegEx here
-# is a helper you can work with.
-# flake8 example:
-# If you use following format argument with flake8 you can use the regex below to parse it.
-# TOOL_ARGS += ["--format='%(row)d,%(col)d,%(code).1s,%(code)s:%(text)s'"]
-# DIAGNOSTIC_RE =
-#    r"(?P<line>\d+),(?P<column>-?\d+),(?P<type>\w+),(?P<code>\w+\d+):(?P<message>[^\r\n]*)"
-DIAGNOSTIC_RE = re.compile(r"")
-
-
-def _parse_output_using_regex(content: str) -> list[lsp.Diagnostic]:
-    lines: list[str] = content.splitlines()
-    diagnostics: list[lsp.Diagnostic] = []
-
-    # TODO: Determine if your linter reports line numbers starting at 1 (True) or 0 (False).
-    line_at_1 = True
-    # TODO: Determine if your linter reports column numbers starting at 1 (True) or 0 (False).
-    column_at_1 = True
-
-    line_offset = 1 if line_at_1 else 0
-    col_offset = 1 if column_at_1 else 0
-    for line in lines:
-        if line.startswith("'") and line.endswith("'"):
-            line = line[1:-1]
-        match = DIAGNOSTIC_RE.match(line)
-        if match:
-            data = match.groupdict()
-            position = lsp.Position(
-                line=max([int(data["line"]) - line_offset, 0]),
-                character=int(data["column"]) - col_offset,
-            )
-            diagnostic = lsp.Diagnostic(
-                range=lsp.Range(
-                    start=position,
-                    end=position,
-                ),
-                message=data.get("message"),
-                severity=_get_severity(data["code"], data["type"]),
-                code=data["code"],
-                source=TOOL_MODULE,
-            )
-            diagnostics.append(diagnostic)
-
-    return diagnostics
-
-
-# TODO: if you want to handle setting specific severity for your linter
-# in a user configurable way, then look at look at how it is implemented
-# for `pylint` extension from our team.
-# Pylint: https://github.com/microsoft/vscode-pylint
-# Follow the flow of severity from the settings in package.json to the server.
-def _get_severity(*_codes: list[str]) -> lsp.DiagnosticSeverity:
-    # TODO: All reported issues from linter are treated as warning.
-    # change it as appropriate for your linter.
-    return lsp.DiagnosticSeverity.Warning
-
-
-# **********************************************************
-# Linting features end here
-# **********************************************************
-
-# TODO: If your tool is a formatter then update this section.
-# Delete "Formatting features" section if your tool is NOT a
-# formatter.
-# **********************************************************
-# Formatting features start here
-# **********************************************************
-#  Sample implementations:
-#  Black: https://github.com/microsoft/vscode-black-formatter/blob/main/bundled/tool
-
+    if result and result.stdout:
+        return _parse_json_output(result.stdout, document.uri)
+    return []
 
-@LSP_SERVER.feature(lsp.TEXT_DOCUMENT_FORMATTING)
-def formatting(params: lsp.DocumentFormattingParams) -> list[lsp.TextEdit] | None:
-    """LSP handler for textDocument/formatting request."""
-    # If your tool is a formatter you can use this handler to provide
-    # formatting support on save. You have to return an array of lsp.TextEdit
-    # objects, to provide your formatted results.
-
-    document = LSP_SERVER.workspace.get_text_document(params.text_document.uri)
-    edits = _formatting_helper(document)
-    if edits:
-        return edits
 
-    # NOTE: If you provide [] array, VS Code will clear the file of all contents.
-    # To indicate no changes to file return None.
-    return None
+def _parse_json_output(content: str, doc_uri: str) -> list[lsp.Diagnostic]:
+    """Parses PythonTA's JSON output and maps it to LSP Diagnostics."""
+    json_start = content.find("[")
+    if json_start == -1:
+        return []
+    content = content[json_start:]
 
+    raw_results = json.loads(content)
+    diagnostics_data = []
+    for file_result in raw_results:
+        file_uri = file_result.get("uri")
+        if file_uri and _normalize_uri_path(file_uri) == _normalize_uri_path(doc_uri):
+            diagnostics_data = file_result.get("diagnostics", [])
+            break
 
-def _formatting_helper(document: workspace.TextDocument) -> list[lsp.TextEdit] | None:
-    # TODO: For formatting on save support the formatter you use must support
-    # formatting via stdin.
-    # Read, and update_run_tool_on_document and _run_tool functions as needed
-    # for your formatter.
-    result = _run_tool_on_document(document, use_stdin=True)
-    if result.stdout:
-        new_source = _match_line_endings(document, result.stdout)
-        return [
-            lsp.TextEdit(
-                range=lsp.Range(
-                    start=lsp.Position(line=0, character=0),
-                    end=lsp.Position(line=len(document.lines), character=0),
-                ),
-                new_text=new_source,
-            )
-        ]
-    return None
+    for diag in diagnostics_data:
+        if "severity" in diag and isinstance(diag["severity"], int):
+            diag["severity"] = lsp.DiagnosticSeverity(diag["severity"])
 
+    converter = converters.get_converter()
+    return converter.structure(diagnostics_data, list[lsp.Diagnostic])
 
-def _get_line_endings(lines: list[str]) -> str:
-    """Returns line endings used in the text."""
-    try:
-        if lines[0][-2:] == "\r\n":
-            return "\r\n"
-        return "\n"
-    except Exception:  # pylint: disable=broad-except
-        return None
 
-
-def _match_line_endings(document: workspace.TextDocument, text: str) -> str:
-    """Ensures that the edited text line endings matches the document line endings."""
-    expected = _get_line_endings(document.source.splitlines(keepends=True))
-    actual = _get_line_endings(text.splitlines(keepends=True))
-    if actual == expected or actual is None or expected is None:
-        return text
-    return text.replace(actual, expected)
-
-
-# **********************************************************
-# Formatting features ends here
-# **********************************************************
+def _normalize_uri_path(uri: str) -> str:
+    """
+    Normalizes a file URI to a comparable filesystem path.
+    On Windows, `pathlib.Path.resolve()` uppercases drive letters, while 
+    VS Code always sends `document.uri` with a lowercase drive letter. 
+    Comparing raw URI strings therefore never matches on Windows.
+    """
+    fs_path = uris.to_fs_path(uri)
+    
+    if fs_path is None:
+        return ""
+        
+    return os.path.normcase(os.path.normpath(fs_path))
 
 
 # **********************************************************

package.json

@@ -64,13 +64,6 @@
         "vsce-package": "vsce package -o python-ta.vsix"
     },
     "contributes": {
-        "keybindings": [
-            {
-                "command": "python-ta.check",
-                "key": "ctrl+shift+t",
-                "when": "editorTextFocus && editorLangId == 'python'"
-            }
-        ],
         "configuration": {
             "properties": {
                 "python-ta.configPath": {
@@ -151,11 +144,6 @@
                 "title": "Restart Server",
                 "category": "PythonTA VS Code Extension",
                 "command": "python-ta.restart"
-            },
-            {
-                "title": "Run PythonTA",
-                "category": "PythonTA VS Code Extension",
-                "command": "python-ta.check"
             }
         ]
     },

pyproject.toml

@@ -6,7 +6,7 @@ readme = "README.md"
 requires-python = ">=3.13"
 dependencies = [
     "pygls>=2.1.1",
-    "python-ta",
+    "python-ta @ git+https://github.com/pyta-uoft/pyta.git#subdirectory=packages/python-ta",
 ]
 
 [dependency-groups]