From 74ac148747574b3060c5935d58e5a669321b07cc Mon Sep 17 00:00:00 2001 From: Dan Helfman Date: Wed, 19 Jun 2019 20:48:54 -0700 Subject: [PATCH] Disable console color via "color" option in borgmatic configuration output section (#191). --- NEWS | 3 ++ borgmatic/commands/borgmatic.py | 70 +++++++++++++++--------- borgmatic/config/schema.yaml | 10 ++++ borgmatic/config/validate.py | 9 ---- borgmatic/logger.py | 8 ++- docs/how-to/set-up-backups.md | 11 ++-- setup.py | 2 +- tests/unit/commands/test_borgmatic.py | 77 +++++++++++++++------------ tests/unit/config/test_validate.py | 20 +------ tests/unit/test_logger.py | 47 ++++++++++++++-- 10 files changed, 155 insertions(+), 102 deletions(-) diff --git a/NEWS b/NEWS index a3bb588..c807762 100644 --- a/NEWS +++ b/NEWS @@ -1,3 +1,6 @@ +1.3.8 + * #191: Disable console color via "color" option in borgmatic configuration output section. + 1.3.7 * #196: Fix for unclear error message for invalid YAML merge include. * #197: Don't color syslog output. diff --git a/borgmatic/commands/borgmatic.py b/borgmatic/commands/borgmatic.py index 6113c3a..26cbdd2 100644 --- a/borgmatic/commands/borgmatic.py +++ b/borgmatic/commands/borgmatic.py @@ -404,37 +404,50 @@ def run_actions( yield json.loads(json_output) -def collect_configuration_run_summary_logs(config_filenames, args): +def load_configurations(config_filenames): ''' - Given a sequence of configuration filenames and parsed command-line arguments as an - argparse.ArgumentParser instance, run each configuration file and yield a series of - logging.LogRecord instances containing summary information about each run. - - As a side effect of running through these configuration files, output their JSON results, if - any, to stdout. + Given a sequence of configuration filenames, load and validate each configuration file. Return + the results as a tuple of: dict of configuration filename to corresponding parsed configuration, + and sequence of logging.LogRecord instances containing any parse errors. ''' # Dict mapping from config filename to corresponding parsed config dict. configs = collections.OrderedDict() + logs = [] # Parse and load each configuration file. for config_filename in config_filenames: try: - logger.debug('{}: Parsing configuration file'.format(config_filename)) configs[config_filename] = validate.parse_configuration( config_filename, validate.schema_filename() ) except (ValueError, OSError, validate.Validation_error) as error: - yield logging.makeLogRecord( - dict( - levelno=logging.CRITICAL, - levelname='CRITICAL', - msg='{}: Error parsing configuration file'.format(config_filename), - ) - ) - yield logging.makeLogRecord( - dict(levelno=logging.CRITICAL, levelname='CRITICAL', msg=error) + logs.extend( + [ + logging.makeLogRecord( + dict( + levelno=logging.CRITICAL, + levelname='CRITICAL', + msg='{}: Error parsing configuration file'.format(config_filename), + ) + ), + logging.makeLogRecord( + dict(levelno=logging.CRITICAL, levelname='CRITICAL', msg=error) + ), + ] ) + return (configs, logs) + + +def collect_configuration_run_summary_logs(configs, args): + ''' + Given a dict of configuration filename to corresponding parsed configuration, and parsed + command-line arguments as an argparse.ArgumentParser instance, run each configuration file and + yield a series of logging.LogRecord instances containing summary information about each run. + + As a side effect of running through these configuration files, output their JSON results, if + any, to stdout. + ''' # Run cross-file validation checks. if args.extract or (args.list and args.archive): try: @@ -472,7 +485,7 @@ def collect_configuration_run_summary_logs(config_filenames, args): if json_results: sys.stdout.write(json.dumps(json_results)) - if not config_filenames: + if not configs: yield logging.makeLogRecord( dict( levelno=logging.CRITICAL, @@ -507,25 +520,30 @@ def main(): # pragma: no cover logger.critical('Error parsing arguments: {}'.format(' '.join(sys.argv))) exit_with_help_link() - colorama.init(autoreset=True, strip=not should_do_markup(args.no_color)) - - configure_logging( - verbosity_to_log_level(args.verbosity), verbosity_to_log_level(args.syslog_verbosity) - ) - if args.version: print(pkg_resources.require('borgmatic')[0].version) sys.exit(0) config_filenames = tuple(collect.collect_config_filenames(args.config_paths)) + configs, parse_logs = load_configurations(config_filenames) + + colorama.init(autoreset=True, strip=not should_do_markup(args.no_color, configs)) + configure_logging( + verbosity_to_log_level(args.verbosity), verbosity_to_log_level(args.syslog_verbosity) + ) + logger.debug('Ensuring legacy configuration is upgraded') convert.guard_configuration_upgraded(LEGACY_CONFIG_PATH, config_filenames) - summary_logs = tuple(collect_configuration_run_summary_logs(config_filenames, args)) + summary_logs = list(collect_configuration_run_summary_logs(configs, args)) logger.info('') logger.info('summary:') - [logger.handle(log) for log in summary_logs if log.levelno >= logger.getEffectiveLevel()] + [ + logger.handle(log) + for log in parse_logs + summary_logs + if log.levelno >= logger.getEffectiveLevel() + ] if any(log.levelno == logging.CRITICAL for log in summary_logs): exit_with_help_link() diff --git a/borgmatic/config/schema.yaml b/borgmatic/config/schema.yaml index 0aca248..0c5c62e 100644 --- a/borgmatic/config/schema.yaml +++ b/borgmatic/config/schema.yaml @@ -301,6 +301,16 @@ map: this prefix. Borg placeholders can be used. See the output of "borg help placeholders" for details. Defaults to "{hostname}-". example: sourcehostname + output: + desc: | + Options for customizing borgmatic's own output and logging. + map: + color: + type: bool + desc: | + Apply color to console output. Can be overridden with --no-color command-line + flag. Defaults to true. + example: false hooks: desc: | Shell commands or scripts to execute before and after a backup or if an error has occurred. diff --git a/borgmatic/config/validate.py b/borgmatic/config/validate.py index 5a28633..02ab704 100644 --- a/borgmatic/config/validate.py +++ b/borgmatic/config/validate.py @@ -7,8 +7,6 @@ import ruamel.yaml from borgmatic.config import load -logger = logging.getLogger(__name__) - def schema_filename(): ''' @@ -65,13 +63,6 @@ def apply_logical_validation(config_filename, parsed_configuration): ), ) - consistency_prefix = parsed_configuration.get('consistency', {}).get('prefix') - if archive_name_format and not consistency_prefix: - logger.warning( - 'Since version 1.1.16, if you provide `archive_name_format`, you should also' - ' specify `consistency.prefix`.' - ) - def parse_configuration(config_filename, schema_filename): ''' diff --git a/borgmatic/logger.py b/borgmatic/logger.py index 439f09f..eb00584 100644 --- a/borgmatic/logger.py +++ b/borgmatic/logger.py @@ -21,13 +21,17 @@ def to_bool(arg): return False -def should_do_markup(no_color): +def should_do_markup(no_color, configs): ''' - Determine if we should enable colorama marking up. + Given the value of the command-line no-color argument, and a dict of configuration filename to + corresponding parsed configuration, determine if we should enable colorama marking up. ''' if no_color: return False + if any(config.get('output', {}).get('color') is False for config in configs.values()): + return False + py_colors = os.environ.get('PY_COLORS', None) if py_colors is not None: diff --git a/docs/how-to/set-up-backups.md b/docs/how-to/set-up-backups.md index bb368bc..1bce273 100644 --- a/docs/how-to/set-up-backups.md +++ b/docs/how-to/set-up-backups.md @@ -199,12 +199,13 @@ sudo systemctl start borgmatic.timer Feel free to modify the timer file based on how frequently you'd like borgmatic to run. -## Colored Output +## Colored output -Borgmatic uses [colorama](https://pypi.org/project/colorama/) to produce -colored terminal output by default. It is disabled when a non-interactive -terminal is detected (like a cron job). Otherwise, it can be disabled by -passing `--no-color` or by setting the environment variable `PY_COLORS=False`. +Borgmatic produces colored terminal output by default. It is disabled when a +non-interactive terminal is detected (like a cron job). Otherwise, you can +disable it by passing the `--no-color` flag, setting the environment variable +`PY_COLORS=False`, or setting the `color` option to `false` in the `output` +section of configuration. ## Troubleshooting diff --git a/setup.py b/setup.py index 55ea48d..8fa6485 100644 --- a/setup.py +++ b/setup.py @@ -1,6 +1,6 @@ from setuptools import find_packages, setup -VERSION = '1.3.7' +VERSION = '1.3.8' setup( diff --git a/tests/unit/commands/test_borgmatic.py b/tests/unit/commands/test_borgmatic.py index e561087..ced1624 100644 --- a/tests/unit/commands/test_borgmatic.py +++ b/tests/unit/commands/test_borgmatic.py @@ -3,99 +3,106 @@ from flexmock import flexmock from borgmatic.commands import borgmatic as module +def test_load_configurations_collects_parsed_configurations(): + configuration = flexmock() + other_configuration = flexmock() + flexmock(module.validate).should_receive('parse_configuration').and_return( + configuration + ).and_return(other_configuration) + + configs, logs = tuple(module.load_configurations(('test.yaml', 'other.yaml'))) + + assert configs == {'test.yaml': configuration, 'other.yaml': other_configuration} + assert logs == [] + + +def test_load_configurations_logs_critical_for_parse_error(): + flexmock(module.validate).should_receive('parse_configuration').and_raise(ValueError) + + configs, logs = tuple(module.load_configurations(('test.yaml',))) + + assert configs == {} + assert any(log for log in logs if log.levelno == module.logging.CRITICAL) + + def test_collect_configuration_run_summary_logs_info_for_success(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) flexmock(module).should_receive('run_configuration').and_return([]) args = flexmock(extract=False, list=False) - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) + logs = tuple(module.collect_configuration_run_summary_logs({'test.yaml': {}}, args=args)) assert all(log for log in logs if log.levelno == module.logging.INFO) def test_collect_configuration_run_summary_logs_info_for_success_with_extract(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) flexmock(module.validate).should_receive('guard_configuration_contains_repository') flexmock(module).should_receive('run_configuration').and_return([]) args = flexmock(extract=True, list=False, repository='repo') - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) + logs = tuple(module.collect_configuration_run_summary_logs({'test.yaml': {}}, args=args)) assert all(log for log in logs if log.levelno == module.logging.INFO) def test_collect_configuration_run_summary_logs_critical_for_extract_with_repository_error(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) flexmock(module.validate).should_receive('guard_configuration_contains_repository').and_raise( ValueError ) args = flexmock(extract=True, list=False, repository='repo') - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) + logs = tuple(module.collect_configuration_run_summary_logs({'test.yaml': {}}, args=args)) assert any(log for log in logs if log.levelno == module.logging.CRITICAL) def test_collect_configuration_run_summary_logs_critical_for_list_with_archive_and_repository_error(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) flexmock(module.validate).should_receive('guard_configuration_contains_repository').and_raise( ValueError ) args = flexmock(extract=False, list=True, repository='repo', archive='test') - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) + logs = tuple(module.collect_configuration_run_summary_logs({'test.yaml': {}}, args=args)) assert any(log for log in logs if log.levelno == module.logging.CRITICAL) def test_collect_configuration_run_summary_logs_info_for_success_with_list(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) flexmock(module).should_receive('run_configuration').and_return([]) args = flexmock(extract=False, list=True, repository='repo', archive=None) - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) + logs = tuple(module.collect_configuration_run_summary_logs({'test.yaml': {}}, args=args)) assert all(log for log in logs if log.levelno == module.logging.INFO) -def test_collect_configuration_run_summary_logs_critical_for_parse_error(): - flexmock(module.validate).should_receive('parse_configuration').and_raise(ValueError) - args = flexmock(extract=False, list=False) - - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) - - assert any(log for log in logs if log.levelno == module.logging.CRITICAL) - - def test_collect_configuration_run_summary_logs_critical_for_run_error(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) flexmock(module.validate).should_receive('guard_configuration_contains_repository') flexmock(module).should_receive('run_configuration').and_raise(ValueError) args = flexmock(extract=False, list=False) - logs = tuple(module.collect_configuration_run_summary_logs(('test.yaml',), args=args)) - - assert any(log for log in logs if log.levelno == module.logging.CRITICAL) - - -def test_collect_configuration_run_summary_logs_critical_for_missing_configs(): - flexmock(module.validate).should_receive('parse_configuration').and_return({'test.yaml': {}}) - flexmock(module).should_receive('run_configuration').and_return([]) - args = flexmock(config_paths=(), extract=False, list=False) - - logs = tuple(module.collect_configuration_run_summary_logs(config_filenames=(), args=args)) + logs = tuple(module.collect_configuration_run_summary_logs({'test.yaml': {}}, args=args)) assert any(log for log in logs if log.levelno == module.logging.CRITICAL) def test_collect_configuration_run_summary_logs_outputs_merged_json_results(): - flexmock(module.validate).should_receive('parse_configuration').and_return( - {'test.yaml': {}, 'test2.yaml': {}} - ) flexmock(module).should_receive('run_configuration').and_return(['foo', 'bar']).and_return( ['baz'] ) flexmock(module.sys.stdout).should_receive('write').with_args('["foo", "bar", "baz"]').once() args = flexmock(extract=False, list=False) - tuple(module.collect_configuration_run_summary_logs(('test.yaml', 'test2.yaml'), args=args)) + tuple( + module.collect_configuration_run_summary_logs( + {'test.yaml': {}, 'test2.yaml': {}}, args=args + ) + ) + + +def test_collect_configuration_run_summary_logs_critical_for_missing_configs(): + flexmock(module).should_receive('run_configuration').and_return([]) + args = flexmock(extract=False, list=False, config_paths=[]) + + logs = tuple(module.collect_configuration_run_summary_logs({}, args=args)) + + assert any(log for log in logs if log.levelno == module.logging.CRITICAL) diff --git a/tests/unit/config/test_validate.py b/tests/unit/config/test_validate.py index c31d969..99c489a 100644 --- a/tests/unit/config/test_validate.py +++ b/tests/unit/config/test_validate.py @@ -1,5 +1,4 @@ import pytest -from flexmock import flexmock from borgmatic.config import validate as module @@ -37,20 +36,6 @@ def test_apply_logical_validation_raises_if_archive_name_format_present_without_ ) -def test_apply_logical_validation_warns_if_archive_name_format_present_without_consistency_prefix(): - logger = flexmock(module.logger) - logger.should_receive('warning').once() - - module.apply_logical_validation( - 'config.yaml', - { - 'storage': {'archive_name_format': '{hostname}-{now}'}, - 'retention': {'prefix': '{hostname}-'}, - 'consistency': {}, - }, - ) - - def test_apply_locical_validation_raises_if_unknown_repository_in_check_repositories(): with pytest.raises(module.Validation_error): module.apply_logical_validation( @@ -74,10 +59,7 @@ def test_apply_locical_validation_does_not_raise_if_known_repository_in_check_re ) -def test_apply_logical_validation_does_not_raise_or_warn_if_archive_name_format_and_prefix_present(): - logger = flexmock(module.logger) - logger.should_receive('warning').never() - +def test_apply_logical_validation_does_not_raise_if_archive_name_format_and_prefix_present(): module.apply_logical_validation( 'config.yaml', { diff --git a/tests/unit/test_logger.py b/tests/unit/test_logger.py index c1e538b..866a7cc 100644 --- a/tests/unit/test_logger.py +++ b/tests/unit/test_logger.py @@ -21,34 +21,71 @@ def test_to_bool_passes_none_through(): def test_should_do_markup_respects_no_color_value(): - assert module.should_do_markup(no_color=True) is False + assert module.should_do_markup(no_color=True, configs={}) is False + + +def test_should_do_markup_respects_config_value(): + assert ( + module.should_do_markup(no_color=False, configs={'foo.yaml': {'output': {'color': False}}}) + is False + ) + + +def test_should_do_markup_prefers_any_false_config_value(): + assert ( + module.should_do_markup( + no_color=False, + configs={ + 'foo.yaml': {'output': {'color': True}}, + 'bar.yaml': {'output': {'color': False}}, + }, + ) + is False + ) def test_should_do_markup_respects_PY_COLORS_environment_variable(): flexmock(module.os.environ).should_receive('get').and_return('True') flexmock(module).should_receive('to_bool').and_return(True) - assert module.should_do_markup(no_color=False) is True + assert module.should_do_markup(no_color=False, configs={}) is True + + +def test_should_do_markup_prefers_no_color_value_to_config_value(): + assert ( + module.should_do_markup(no_color=True, configs={'foo.yaml': {'output': {'color': True}}}) + is False + ) + + +def test_should_do_markup_prefers_config_value_to_PY_COLORS(): + flexmock(module.os.environ).should_receive('get').and_return('True') + flexmock(module).should_receive('to_bool').and_return(True) + + assert ( + module.should_do_markup(no_color=False, configs={'foo.yaml': {'output': {'color': False}}}) + is False + ) def test_should_do_markup_prefers_no_color_value_to_PY_COLORS(): flexmock(module.os.environ).should_receive('get').and_return('True') flexmock(module).should_receive('to_bool').and_return(True) - assert module.should_do_markup(no_color=True) is False + assert module.should_do_markup(no_color=True, configs={}) is False def test_should_do_markup_respects_stdout_tty_value(): flexmock(module.os.environ).should_receive('get').and_return(None) - assert module.should_do_markup(no_color=False) is False + assert module.should_do_markup(no_color=False, configs={}) is False def test_should_do_markup_prefers_PY_COLORS_to_stdout_tty_value(): flexmock(module.os.environ).should_receive('get').and_return('True') flexmock(module).should_receive('to_bool').and_return(True) - assert module.should_do_markup(no_color=False) is True + assert module.should_do_markup(no_color=False, configs={}) is True def test_console_color_formatter_format_includes_log_message():