libenvpp - Modern C++ Library for Handling Environment Variables

This library provides a modern, platform independent, type-safe way of handling environment variables. It offers an easy to use interface for the most common use-cases, and is extensible and customizable to more complex use-cases, if and when required.

Features
Usage
Error Handling
Testing
Installation

Features

Platform independent
Type-safe parsing of environment variables
Support for required and optional environment variables, with specifiable default value
Automatic parsing of built-in types
User-defined types parsable with user-defined parser
Optional user-defined validation
Convenience range/option environment variable type
Parsing/validating possible per type, or per environment variable
Typo detection based on edit-distance
Unused environment variable detection

Usage

Simple Example

To use the library the first step would be to include the header:

#include <libenvpp/env.hpp>

The library is built around the idea of a prefix. The assumption is that all relevant environment variables used by a program share the same prefix, for example with the prefix MYPROG:

MYPROG_LOG_FILE_PATH="/path/to/log"
MYPROG_NUM_THREADS=4

This would be expressed as:

auto pre = env::prefix("MYPROG");

const auto log_path_id = pre.register_variable<std::filesystem::path>("LOG_FILE_PATH");
const auto num_threads_id = pre.register_required_variable<unsigned int>("NUM_THREADS");

Here MYPROG_LOG_FILE_PATH is registered as an optional environment variable, which will not cause an error or warning when it is not specified in the environment. Whereas MYPROG_NUM_THREADS is registered as a required environment variable, which will cause an error if it is not found in the environment.

After having registered all variables with the corresponding prefix it can be parsed and validated, which will retrieve the values from the environment and parse them into the type specified when registering:

const auto parsed_and_validated_pre = pre.parse_and_validate();

This will invalidate the prefix pre which must not be used after this point, and can be safely destroyed.

If anything went wrong, like the required variable not being found, or the variable not being parsable, this can be queried through the parsed_and_validated_pre:

if (!parsed_and_validated_pre.ok()) {
    std::cout << parsed_and_validated_pre.warning_message();
    std::cout << parsed_and_validated_pre.error_message();
}

Otherwise the parsed values can be retrieved from the parsed_and_validated_pre:

if (parsed_and_validated_pre.ok()) {
    const auto log_path = parsed_and_validated_pre.get_or(log_path_id, "/default/log/path");
    const auto num_threads = parsed_and_validated_pre.get(num_threads_id);
}

Optional variables can be given a default value when getting them with get_or, which will return the default value if (and only if) the variable was not found in the environment. Parsing or validation errors of optional variables are also reported as errors. Additionally, optional variables can also be retrieved with get which will return a std::optional which is empty if the variable was not found in the environment.

Required variables can only be gotten with get, as they are required and therefore need no default value, which will return the type directly. If required variables were not found in the environment this will be reported as an error.

Simple Example - Code

Putting everything together:

#include <cstdlib>
#include <filesystem>
#include <iostream>

#include <libenvpp/env.hpp>

int main()
{
    auto pre = env::prefix("MYPROG");

    const auto log_path_id = pre.register_variable<std::filesystem::path>("LOG_FILE_PATH");
    const auto num_threads_id = pre.register_required_variable<unsigned int>("NUM_THREADS");

    const auto parsed_and_validated_pre = pre.parse_and_validate();

    if (parsed_and_validated_pre.ok()) {
        const auto log_path = parsed_and_validated_pre.get_or(log_path_id, "/default/log/path");
        const auto num_threads = parsed_and_validated_pre.get(num_threads_id);

        std::cout << "Log path   : " << log_path << std::endl;
        std::cout << "Num threads: " << num_threads << std::endl;
    } else {
        std::cout << parsed_and_validated_pre.warning_message();
        std::cout << parsed_and_validated_pre.error_message();
    }

    return EXIT_SUCCESS;
}

Simple Example - Output

Running the example without having any environment variable set:

$ ./libenvpp_simple_usage_example
Error  : Environment variable 'MYPROG_NUM_THREADS' not set

Running the example with a typo:

$ MYPROG_LOG_FILEPATH=/var/log/file ./libenvpp_simple_usage_example
Warning: Unrecognized environment variable 'MYPROG_LOG_FILEPATH' set, did you mean 'MYPROG_LOG_FILE_PATH'?
Error  : Environment variable 'MYPROG_NUM_THREADS' not set

Running the example with only the required variable set:

$ MYPROG_NUM_THREADS=4 ./libenvpp_simple_usage_example
Log path   : "/default/log/path"
Num threads: 4

Running the example with both variables set:

$ MYPROG_LOG_FILE_PATH=/var/log/file MYPROG_NUM_THREADS=4 ./libenvpp_simple_usage_example
Log path   : "/var/log/file"
Num threads: 4

Custom Type Parser

To provide a parser for a user-defined type it is necessary to specialize the template struct env::default_parser for the specific type that should be parsed and provide a call operator that takes a std::string_view and returns the parsed type, for example:

struct program_data {
    int number;
    float percent;
};

namespace env {
template <>
struct default_parser<program_data> {
    program_data operator()(const std::string_view str) const
    {
        const auto split_str = split(str, ',');
        if (split_str.size() != 2) {
            // Report an error if the input does not have the expected format
            throw parser_error{"Expected 2 comma delimited values"};
        }

        auto parsed_data = program_data{};

        // Delegate parsing of primitive types to the default_parser
        parsed_data.number = default_parser<int>{}(split_str[0]);
        parsed_data.percent = default_parser<float>{}(split_str[1]);

        return parsed_data;
    }
};
} // namespace env

Note: The default_parser already supports primitive types (and everything that can be constructed from string), so parsing should be delegated to the existing implementation whenever possible.

Custom Type Parser - Code

For the entire code see examples/libenvpp_custom_parser_example.cpp.

Custom Type Validator

Similarly to the env::default_parser there is also an env::default_validator which can be specialized for any type that should be validated. The non-specialized implementation of default_validator does nothing, as everything that can be parsed is considered valid by default. When specializing default_validator for a type the struct must contain a call operator that takes the parsed type and returns void. Validation errors should be reported by throwing an env::validation_error:

namespace env {
template <>
struct default_validator<std::filesystem::path> {
    void operator()(const std::filesystem::path& path) const
    {
        if (!std::filesystem::exists(path)) {
            throw validation_error{path.string() + " path does not exist"};
        }
    }
};
} // namespace env

The validator example above shows a validator for the type std::filesystem::path which requires that environment variables of that type must exist on the filesystem, otherwise the specified environment variable is considered invalid, even if the string was a valid path.

Custom Type Validator - Code

For the full example see examples/libenvpp_custom_validator_example.cpp.

Custom Variable Parser and Validator

If parsing/validator should be done in a specific way for a specific variable only, and not for every variable of the same type, it is possible to specify a parser and validator function when registering a variable:

std::filesystem::path path_parser_and_validator(const std::string_view str)
{
    const auto log_path = std::filesystem::path(str);

    if (!std::filesystem::exists(log_path)) {
        if (!std::filesystem::create_directory(log_path)) {
            throw env::validation_error{"Unable to create log directory"};
        }
    } else if (!std::filesystem::is_directory(log_path)) {
        throw env::validation_error{"Log path is not a directory"};
    }

    return log_path;
}

int main()
{
    auto pre = env::prefix("CUSTOM_PARSER_AND_VALIDATOR");

    const auto path_id = pre.register_required_variable<std::filesystem::path>("LOG_PATH", path_parser_and_validator);

    /*...*/
}

For example, the parser and validator function above will make sure that the log path specified by the environment variable points to a directory, and will even create it if it doesn't exist. Errors are reported by throwing env::parser_error or env::validation_error.

Note: When specifying a custom parser and validator function, the default_parser and default_validator for the given type are not invoked automatically, however delegating to them is still possible and should be done if appropriate.

Note: The parser and validator function passed to register_[required]_variable can be any callable type, be it function, lambda, or functor.

Custom Variable Parser and Validator - Code

For the full example see examples/libenvpp_custom_parser_and_validator_example.cpp.

Range Variables

Because it is a frequent use-case that a value must be within a given range, environment variables can be registered with register_range which additionally takes a minimum and maximum value, and validates that the parsed value is within the given range. The minimum and maximum values are both inclusive. For example:

auto pre = env::prefix("RANGE");
const auto num_threads_id = pre.register_range<unsigned int>("NUM_THREADS", 1, std::thread::hardware_concurrency());
const auto parsed_and_validated_pre = pre.parse_and_validate();

if (parsed_and_validated_pre.ok()) {
    const auto num_threads = parsed_and_validated_pre.get_or(num_threads_id, std::thread::hardware_concurrency());
    std::cout << "Number of threads: " << num_threads << std::endl;
} else {
    std::cout << parsed_and_validated_pre.warning_message();
    std::cout << parsed_and_validated_pre.error_message();
}

This will enforce that the specified number of threads is between 1 and the maximum that the system this is run on supports, inclusive.

Note: Range variables, just like normal variables, can be optional or required. And the type is parsed with default_parser and validated with default_validator. It is not possible, however, to specify a custom parser and validator function for a range, if this is desired, a normal variable should be used and the range check done manually as part of the validation.

Note: When using get_or to set a default value for an optional range, the range requirement is not enforced on the default. This can be used for special case handling if desired.

Range Variables - Code

For the complete code see examples/libenvpp_range_example.cpp.

Option Variables

Another frequent use-case is that a value is one of a given set of options. For this an environment variable can be registered with register_[required]_option, which takes a list of valid options, against which the value is checked. For example:

enum class option {
    first_choice,
    second_choice,
    third_choice,
    default_choice,
};

int main()
{
    auto pre = env::prefix("OPTION");

    const auto option_id =
        pre.register_option<option>("CHOICE", {option::first_choice, option::second_choice, option::third_choice});

    const auto parsed_and_validated_pre = pre.parse_and_validate();

    if (parsed_and_validated_pre.ok()) {
        const auto opt = parsed_and_validated_pre.get_or(option_id, option::default_choice);
    }
}

This registers an enum class option, where only a subset of all possible values is considered valid, so that option::default_choice can be used as the value if the variable is not set.

Note: The list of options provided when registering must not be empty, and must not contain duplicates.

Note: As with range variables, the default value given with get_or is not enforced to be within the list of options given when registering the option variable.

Note: Since C++ does not provide any way to automatically parse enum class types from string, the example above additionally requires a specialized default_parser for the enum class type.

Note: Options are mostly intended to be used with enum class types, but this is in no way a requirement. Any type can be used as an option, and enum class types can also just be normal environment variables.

Option Variables - Code

For the full code, including the parser for the enum class, see examples/libenvpp_option_example.cpp.

Prefixless Environment Variables

Even though it is recommended to namespace environment variables with a prefix, and use the prefix mechanism of this library to parse those variables, sometimes it might be necessary to parse environment variables that don't have a prefix. To this end, this library also provides a mechanism for that:

#include <cstdlib>
#include <filesystem>
#include <iostream>

#include <libenvpp/env.hpp>

int main()
{
    const auto log_path = env::get_or<std::filesystem::path>("LOG_FILE_PATH", "/default/log/path");
    const auto num_threads = env::get<unsigned int>("NUM_THREADS");

    if (num_threads.has_value()) {
        std::cout << "Log path   : " << log_path << std::endl;
        std::cout << "Num threads: " << num_threads.value() << std::endl;
    } else {
        std::cout << num_threads.error().what() << std::endl;
    }

    return EXIT_SUCCESS;
}

This is analogous to the simple usage example using a prefix. Required variables can be parsed using env::get, which either returns the parsed and validated value, or an error.

Optional variables can be parsed and validated with env::get_or which will always return a value, either the successfully parsed and validated one from the environment, or the default value.

Note: These functions make use of the same default_parser/default_validator mechanism as the prefix, and so parsing/validating of user-defined types is also supported.

Prefixless Environment Variables - Code

For the code of this example, see examples/libenvpp_prefixless_get_example.cpp.

Error Handling

Help Message

After having registered all variables with a prefix it is possible to generate a formatted help message for a specific prefix. This can be done by calling help_message on either a prefix, or a validated prefix. For example:

auto pre = env::prefix("MYPROG");

const auto log_path_id = pre.register_variable<std::filesystem::path>("LOG_FILE_PATH");
const auto num_threads_id = pre.register_required_variable<unsigned int>("NUM_THREADS");

std::cout << pre.help_message();

will output:

Prefix 'MYPROG' supports the following 2 environment variable(s):
    'MYPROG_LOG_FILE_PATH' optional
    'MYPROG_NUM_THREADS' required

Warnings and Errors

If anything goes wrong when parsing and validating a prefix this can be queried through the following functions:

Function	Return value
`ok()`	`bool` indicating whether there are any errors or warnings
`warning_message()`	`std::string` with a formatted message containing all warnings
`error_message()`	`std::string` with a formatted message containing all errors
`warnings()`	`std::vector<env::error>` containing all warnings
`errors()`	`std::vector<env::error>` containing all errors

`error` Type

The env::error type is used to give more information on warnings and errors. The type supports:

Function	Return value
`get_id()`	`std::size_t` ID of the variable that caused the error/warning. This ID can be compared to the ID returned from the `register_*` functions.
`get_name()`	`std::string` containing the name of the variable that caused the warning/error.
`what()`	`std::string` containing the warning/error message.

Warnings

The following situations will generate a warning:

An optional variable is not set, but a similar (within edit distance configured for the prefix) unused variable is.
An unused variable with the same prefix is set.

Errors

The following situations will generate an error:

A variable could not be parsed.
A variable could not be validated.
The value of a range variable is outside the specified range.
The value of an option variable is not a valid option.
An unexpected exception was thrown while parsing/validating a variable.
A required variable is not set, but a similar (within edit distance configured for the prefix) unused variable is.
A required variable is not set.

Warnings and Errors - Code

For an example of how the warning/error handling functions can be used, see examples/libenvpp_error_handling_example.cpp.

Testing

There are a few ways that help facilitate (unit) testing.

Global Testing Environment

Libenvpp provides an internal global testing environment where variables can be added as key/value pairs using env::scoped_test_environment:

const auto _ = env::scoped_test_environment(std::unordered_map<std::string, std::string>{
    {"MYPROG_LOG_FILE_PATH", "/dev/null"},
    {"MYPROG_NUM_THREADS", "8"},
});

Once the scoped test environment instance goes out of scope, the variables it added to the global test environment are automatically removed.

Any libenvpp function that usually retrieves variables from the system environment will first check if the global test environment contains a value for the requested variable and retrieve that instead. This can be used to facilitate unit testing without having to change the usage of libenvpp.

Note: Any value set in the global testing environment will take precedence over system/custom environment variables.

Note: Interacting with the global testing environment is not thread safe and the user must take care to synchronize any potentially conflicting accesses.

Global Testing Environment - Code

A complete example of how to use the global testing environment, see examples/libenvpp_testing_example.cpp.

Custom Environment

Another mechanism that can be used for testing is the ability to pass a custom environment to prefix::parse_and_validate as the first parameter of type std::unordered_map<std::string, std::string>.

Environment variables will then be fetched from there instead of the system environment. Note that the global testing environment will still take precedence over the custom environment and if the variable is not found in the testing environment or the custom environment no fallback to the system environment will be performed.

Custom Environment - Code

A complete example of how to use a custom environment can be found here: examples/libenvpp_custom_environment_example.cpp

Set for Testing

Additionally it is also possible to bypass parsing and validating completely and directly set the value for a registered variable on a prefix.

auto pre = env::prefix("MYPROG");

const auto log_path_id = pre.register_variable<std::filesystem::path>("LOG_FILE_PATH");
const auto num_threads_id = pre.register_required_variable<unsigned int>("NUM_THREADS");

pre.set_for_testing(log_path_id, "/dev/null");
pre.set_for_testing(num_threads_id, 8);

This will ignore any value for that variable in the testing/system/custom environment and not perform any parsing or validating and just directly return the set value when using get/get_or on the parsed and validated prefix.

Set for Testing - Code

A complete example of how to use set_for_testing can be found here: examples/libenvpp_set_for_testing_example.cpp.

Installation

FetchContent

Libenvpp can be integrated into a cmake project (myproject) using the FetchContent mechanism like this:

CMakeLists.txt:

set(LIBENVPP_INSTALL ON CACHE BOOL "" FORCE) # If installation is desired.
FetchContent_Declare(libenvpp
    GIT_REPOSITORY https://github.com/ph3at/libenvpp.git
    GIT_TAG v1.0.0
)
FetchContent_MakeAvailable(libenvpp)

# ...

target_link_libraries(myproject PUBLIC
    libenvpp::libenvpp
)

myproject-config.cmake.in (only required if installation is desired):

# ...

include(CMakeFindDependencyMacro)
find_dependency(libenvpp REQUIRED)

# ...

Submodule

The library uses cmake as the build system and can be easily integrated into an existing cmake project by adding it with add_subdirectory, ideally used in combination with a git submodule of the library. The build system will detect this use-case and automatically disable the building of examples and tests.

Cmake Options

Option	Default	Description
`LIBENVPP_EXAMPLES`	`ON` if configured standalone, `OFF` if used as dependency	Enables building of example programs.
`LIBENVPP_TESTS`	`ON` if configured standalone, `OFF` if used as dependency	Enables building of unit tests.
`LIBENVPP_CHECKS`	`ON` in debug builds or when tests are enabled, `OFF` in release builds	Custom assertions that are not tied to `NDEBUG`, and are testable by throwing an exception instead of `std::abort`ing iff `LIBENVPP_TESTS` are enabled.
`LIBENVPP_INSTALL`	`OFF`	Adds an install target that can be used to install libenvpp as a library.

Name		Name	Last commit message	Last commit date
Latest commit History 170 Commits
.github/workflows		.github/workflows
cmake		cmake
examples		examples
include/libenvpp		include/libenvpp
source		source
test		test
.clang-format		.clang-format
.editorconfig		.editorconfig
.gitattributes		.gitattributes
.gitignore		.gitignore
CMakeLists.txt		CMakeLists.txt
LICENSE		LICENSE
NOTICE		NOTICE
README.md		README.md

License

cdbhnd/libenvpp

Folders and files

Latest commit

History

Repository files navigation