New CLI subcommand to create clang-compatible compilation database (`compile_commands.json`) (#14370)

* pulled source from dev branch

* missed a file from origin

* formatting

* revised argument names. relaxed matching rules to work for avr too

* add docstrings

* added docs. tightened up regex

* remove unused imports

* cleaning up command file. use existing qmk dir constant

* rename parser library file

* move lib functions into command file. there are only 2 and they aren't large

* currently debugging...

* more robustly find config

* updated docs

* remove unused imports

* reuse make executable from the main make command

* pulled source from dev branch

* missed a file from origin

* formatting

* revised argument names. relaxed matching rules to work for avr too

* add docstrings

* added docs. tightened up regex

* remove unused imports

* cleaning up command file. use existing qmk dir constant

* rename parser library file

* move lib functions into command file. there are only 2 and they aren't large

* currently debugging...

* more robustly find config

* updated docs

* remove unused imports

* reuse make executable from the main make command

* remove MAKEFLAGS from environment for better control over process management

* Update .gitignore

Co-authored-by: Michael Forster <forster@google.com>

* add a usage line to docs

* doc change as suggested

Co-authored-by: Nick Brassel <nick@tzarc.org>

* rename command

* remove debug print statements

* generate-compilation-database: fix arg handling

* generate-comilation-db: improve error handling

* use cli.run() instead of Popen()

Co-authored-by: Xton <cdewan@apple.com>
Co-authored-by: Christon DeWan <cmdpix@mac.com>
Co-authored-by: Michael Forster <forster@google.com>
Co-authored-by: Nick Brassel <nick@tzarc.org>

5 files changed, 166 insertions, 4 deletions

diff --git a/.gitignore b/.gitignore
index f3b76f7edf..768f40b191 100644
--- a/.gitignore
+++ b/.gitignore
@@ -86,3 +86,8 @@ __pycache__
 
 # Allow to exist but don't include it in the repo
 user_song_list.h
+
+# clangd
+compile_commands.json
+.clangd/
+.cache/
diff --git a/docs/cli_commands.md b/docs/cli_commands.md
index 06568afb46..1427b592f5 100644
--- a/docs/cli_commands.md
+++ b/docs/cli_commands.md
@@ -320,6 +320,33 @@ qmk format-c
 qmk format-c -b branch_name
 ```
 
+## `qmk generate-compilation-database`
+
+**Usage**:
+
+```
+qmk generate-compilation-database [-kb KEYBOARD] [-km KEYMAP]
+```
+
+Creates a `compile_commands.json` file.
+
+Does your IDE/editor use a language server but doesn't _quite_ find all the necessary include files? Do you hate red squigglies? Do you wish your editor could figure out `#include QMK_KEYBOARD_H`? You might need a [compilation database](https://clang.llvm.org/docs/JSONCompilationDatabase.html)! The qmk tool can build this for you.
+
+This command needs to know which keyboard and keymap to build. It uses the same configuration options as the `qmk compile` command: arguments, current directory, and config files.
+
+**Example:**
+
+```
+$ cd ~/qmk_firmware/keyboards/gh60/satan/keymaps/colemak
+$ qmk generate-compilation-database
+Ψ Making clean
+Ψ Gathering build instructions from make -n gh60/satan:colemak
+Ψ Found 50 compile commands
+Ψ Writing build database to /Users/you/src/qmk_firmware/compile_commands.json
+```
+
+Now open your dev environment and live a squiggly-free life.
+
 ## `qmk docs`
 
 This command starts a local HTTP server which you can use for browsing or improving the docs. Default port is 8936.
diff --git a/lib/python/qmk/cli/__init__.py b/lib/python/qmk/cli/__init__.py
index 292dcbe812..944938824c 100644
--- a/lib/python/qmk/cli/__init__.py
+++ b/lib/python/qmk/cli/__init__.py
@@ -44,6 +44,7 @@ subcommands = [
     'qmk.cli.format.python',
     'qmk.cli.format.text',
     'qmk.cli.generate.api',
+    'qmk.cli.generate.compilation_database',
     'qmk.cli.generate.config_h',
     'qmk.cli.generate.dfu_header',
     'qmk.cli.generate.docs',
diff --git a/lib/python/qmk/cli/generate/compilation_database.py b/lib/python/qmk/cli/generate/compilation_database.py
new file mode 100755
index 0000000000..2748d96e7a
--- /dev/null
+++ b/lib/python/qmk/cli/generate/compilation_database.py
@@ -0,0 +1,123 @@
+"""Creates a compilation database for the given keyboard build.
+"""
+
+import itertools
+import json
+import os
+import re
+import shlex
+import shutil
+from functools import lru_cache
+from pathlib import Path
+from typing import Dict, Iterator, List, Union
+
+from milc import cli, MILC
+
+from qmk.commands import create_make_command
+from qmk.constants import QMK_FIRMWARE
+from qmk.decorators import automagic_keyboard, automagic_keymap
+
+
+@lru_cache(maxsize=10)
+def system_libs(binary: str) -> List[Path]:
+    """Find the system include directory that the given build tool uses.
+    """
+    cli.log.debug("searching for system library directory for binary: %s", binary)
+    bin_path = shutil.which(binary)
+    return list(Path(bin_path).resolve().parent.parent.glob("*/include")) if bin_path else []
+
+
+file_re = re.compile(r'printf "Compiling: ([^"]+)')
+cmd_re = re.compile(r'LOG=\$\((.+?)&&')
+
+
+def parse_make_n(f: Iterator[str]) -> List[Dict[str, str]]:
+    """parse the output of `make -n <target>`
+
+    This function makes many assumptions about the format of your build log.
+    This happens to work right now for qmk.
+    """
+
+    state = 'start'
+    this_file = None
+    records = []
+    for line in f:
+        if state == 'start':
+            m = file_re.search(line)
+            if m:
+                this_file = m.group(1)
+                state = 'cmd'
+
+        if state == 'cmd':
+            assert this_file
+            m = cmd_re.search(line)
+            if m:
+                # we have a hit!
+                this_cmd = m.group(1)
+                args = shlex.split(this_cmd)
+                args += ['-I%s' % s for s in system_libs(args[0])]
+                new_cmd = ' '.join(shlex.quote(s) for s in args if s != '-mno-thumb-interwork')
+                records.append({"directory": str(QMK_FIRMWARE.resolve()), "command": new_cmd, "file": this_file})
+                state = 'start'
+
+    return records
+
+
+@cli.argument('-kb', '--keyboard', help='The keyboard to build a firmware for. Ignored when a configurator export is supplied.')
+@cli.argument('-km', '--keymap', help='The keymap to build a firmware for. Ignored when a configurator export is supplied.')
+@cli.subcommand('Create a compilation database.')
+@automagic_keyboard
+@automagic_keymap
+def generate_compilation_database(cli: MILC) -> Union[bool, int]:
+    """Creates a compilation database for the given keyboard build.
+
+    Does a make clean, then a make -n for this target and uses the dry-run output to create
+    a compilation database (compile_commands.json). This file can help some IDEs and
+    IDE-like editors work better. For more information about this:
+
+        https://clang.llvm.org/docs/JSONCompilationDatabase.html
+    """
+    command = None
+    # check both config domains: the magic decorator fills in `generate_compilation_database` but the user is
+    # more likely to have set `compile` in their config file.
+    current_keyboard = cli.config.generate_compilation_database.keyboard or cli.config.user.keyboard
+    current_keymap = cli.config.generate_compilation_database.keymap or cli.config.user.keymap
+
+    if current_keyboard and current_keymap:
+        # Generate the make command for a specific keyboard/keymap.
+        command = create_make_command(current_keyboard, current_keymap, dry_run=True)
+    elif not current_keyboard:
+        cli.log.error('Could not determine keyboard!')
+    elif not current_keymap:
+        cli.log.error('Could not determine keymap!')
+
+    if not command:
+        cli.log.error('You must supply both `--keyboard` and `--keymap`, or be in a directory for a keyboard or keymap.')
+        cli.echo('usage: qmk compiledb [-kb KEYBOARD] [-km KEYMAP]')
+        return False
+
+    # remove any environment variable overrides which could trip us up
+    env = os.environ.copy()
+    env.pop("MAKEFLAGS", None)
+
+    # re-use same executable as the main make invocation (might be gmake)
+    clean_command = [command[0], 'clean']
+    cli.log.info('Making clean with {fg_cyan}%s', ' '.join(clean_command))
+    cli.run(clean_command, capture_output=False, check=True, env=env)
+
+    cli.log.info('Gathering build instructions from {fg_cyan}%s', ' '.join(command))
+
+    result = cli.run(command, capture_output=True, check=True, env=env)
+    db = parse_make_n(result.stdout.splitlines())
+    if not db:
+        cli.log.error("Failed to parse output from make output:\n%s", result.stdout)
+        return False
+
+    cli.log.info("Found %s compile commands", len(db))
+
+    dbpath = QMK_FIRMWARE / 'compile_commands.json'
+
+    cli.log.info(f"Writing build database to {dbpath}")
+    dbpath.write_text(json.dumps(db, indent=4))
+
+    return True
diff --git a/lib/python/qmk/commands.py b/lib/python/qmk/commands.py
index 01c23b2612..2995a5fdab 100644
--- a/lib/python/qmk/commands.py
+++ b/lib/python/qmk/commands.py
@@ -28,7 +28,7 @@ def _find_make():
     return make_cmd
 
 
-def create_make_target(target, parallel=1, **env_vars):
+def create_make_target(target, dry_run=False, parallel=1, **env_vars):
     """Create a make command
 
     Args:
@@ -36,6 +36,9 @@ def create_make_target(target, parallel=1, **env_vars):
         target
             Usually a make rule, such as 'clean' or 'all'.
 
+        dry_run
+            make -n -- don't actually build
+
         parallel
             The number of make jobs to run in parallel
 
@@ -52,10 +55,10 @@ def create_make_target(target, parallel=1, **env_vars):
     for key, value in env_vars.items():
         env.append(f'{key}={value}')
 
-    return [make_cmd, *get_make_parallel_args(parallel), *env, target]
+    return [make_cmd, *(['-n'] if dry_run else []), *get_make_parallel_args(parallel), *env, target]
 
 
-def create_make_command(keyboard, keymap, target=None, parallel=1, **env_vars):
+def create_make_command(keyboard, keymap, target=None, dry_run=False, parallel=1, **env_vars):
     """Create a make compile command
 
     Args:
@@ -69,6 +72,9 @@ def create_make_command(keyboard, keymap, target=None, parallel=1, **env_vars):
         target
             Usually a bootloader.
 
+        dry_run
+            make -n -- don't actually build
+
         parallel
             The number of make jobs to run in parallel
 
@@ -84,7 +90,7 @@ def create_make_command(keyboard, keymap, target=None, parallel=1, **env_vars):
     if target:
         make_args.append(target)
 
-    return create_make_target(':'.join(make_args), parallel, **env_vars)
+    return create_make_target(':'.join(make_args), dry_run=dry_run, parallel=parallel, **env_vars)
 
 
 def get_git_version(current_time, repo_dir='.', check_dir='.'):