tscfg is a command line tool that takes a configuration specification parseable by Typesafe Config and generates all the boilerplate to make the definitions available in type-safe, immutable objects (POJOs/records for Java, case classes for Scala).
The generated code only depends on the Typesafe Config library.
- status
- configuration spec
- running tscfg
- configuration access
- supported types
- configuration template
- FAQ
- tests
The tool supports all types handled by Typesafe Config (string, int, long, double, boolean, duration, size-in-bytes, list, object) and has great test coverage. Possible improvements include a more standard command line interface, a proper tscfg library, and perhaps a revision of the syntax for types. Feel free to fork, enter issues/reactions, submit PRs, etc.
In tscfg's approach, the configuration spec itself is any source parseable by Typesafe Config, so the familiar syntax/format and loading mechanisms are used.
For example, from this configuration:
service {
url = "http://example.net/rest"
poolSize = 32
debug = true
factor = 0.75
}
tscfg will generate the following (constructors and other methods omitted):
-
Java:
public class Cfg { public final Service service; public static class Service { public final boolean debug; public final double factor; public final int poolSize; public final String url; } }
Nesting of configuration properties is captured via inner static classes.
-
Scala:
case class Cfg( service : Cfg.Service ) object Cfg { case class Service( debug : Boolean, factor : Double, poolSize : Int, url : String ) }
Nesting of configuration properties is captured via nested companion objects.
The tool determines the type of each field according to the given value in the input configuration. Used in this way, all fields are considered optional, with the given value as the default.
But this wouldn't be flexible enough! To allow the specification of required fields, explicit types, and default values, a string with a simple syntax as follows can be used (illustrated with the integer type):
field spec | meaning | java type / default | scala type / default |
---|---|---|---|
a: "int" | required integer | int / no default |
Int / no default |
a: "int | 3" | optional integer with default value 3 |
int / 3 |
Int / 3 |
a: "int?" | optional integer | Integer / null (*) |
Option[Int] / None |
NOTE
- (*) You can use the
--java:optionals
flag to generateOptional<T>
instead ofnull
.- The type syntax is still subject to change.
The following is a complete example exercising this mechanism.
endpoint {
path: "string"
url: "String | http://example.net"
serial: "int?"
interface {
port: "int | 8080"
}
}
For Java, this basically becomes the immutable class:
public class JavaExampleCfg {
public final Endpoint endpoint;
public static class Endpoint {
public final int intReq;
public final Interface_ interface_;
public final String path;
public final Integer serial;
public final String url;
public static class Interface_ {
public final int port;
public final String type;
}
}
}
And for Scala:
case class ScalaExampleCfg(
endpoint : ScalaExampleCfg.Endpoint
)
object ScalaExampleCfg {
case class Endpoint(
intReq : Int,
interface : Endpoint.Interface,
path : String,
serial : Option[Int],
url : String
)
object Endpoint {
case class Interface(
port : Int,
`type` : Option[String]
)
}
}
You will need a JRE 8+ and the latest fat JAR (tscfg-x.y.z.jar) from the releases.
Or run
sbt assembly
(orsbt ++2.13.7 assembly
) under a clone of this repo to generate the fat jar.
$ java -jar tscfg-x.y.z.jar
tscfg x.y.z
Usage: tscfg.Main --spec inputFile [options]
Options (default):
--pn <packageName> (tscfg.example)
--cn <className> (ExampleCfg)
--dd <destDir> (/tmp if existing or OS dependent temp dir)
--java generate java code (the default)
--java:getters generate getters (see #31) (false)
--java:records generate records (false)
--java:optionals use optionals (false)
--scala generate scala code (java)
--scala:bt use backticks (see #30) (false)
--durations use java.time.Duration (false)
--all-required assume all properties are required (see #47)
--tpl <filename> generate config template (no default)
--tpl.ind <string> template indentation string (" ")
--tpl.cp <string> prefix for template comments ("##")
--withoutTimestamp generate header w/out timestamp (false)
Output is written to $destDir/$className.ext
So, to generate the Java class tscfg.example.ExampleCfg
with the example above
saved in a file example.spec.conf
, you can run:
$ java -jar tscfg-x.y.z.jar --spec example.spec.conf
parsing: example.spec.conf
generating: /tmp/ExampleCfg.java
Please see tscfg-maven-plugin. Thanks @timvlaer!
Please see tscfg-plugin-gradle.
Access to a configuration instance is via usual Typesafe Config mechanism as appropriate for your application, for example, to load the default configuration:
Config tsConfig = ConfigFactory.load().resolve()
or from a given file:
Config tsConfig = ConfigFactory.parseFile(new File("my.conf")).resolve();
Now, to access the configuration fields, instead of, for example:
Config endpoint = tsConfig.getConfig("endpoint");
String path = endpoint.getString("path");
Integer serial = endpoint.hasPathOrNull("serial") ? endpoint.getInt("serial") : null;
int port = endpoint.hasPathOrNull("port") ? endpoint.getInt("interface.port") : 8080;
you can:
-
Create the tscfg generated wrapper:
ExampleCfg cfg = new ExampleCfg(tsConfig);
which will make all verifications about required settings and associated types. In particular, as is typical with Config use, an exception will be thrown if this verification fails.
-
Then, while enjoying full type safety and the code completion and navigation capabilities of your editor or IDE:
String path = cfg.endpoint.path; Integer serial = cfg.endpoint.serial; int port = cfg.endpoint.interface_.port;
An object reference will never be null
(or Optional.empty()
) (None
in Scala) if the corresponding field is required according to
the specification. It will only be null
(or Optional.empty()
) (None
in Scala) if it is marked optional with no default value and
has been omitted in the input configuration.
With this example spec, the generated Java code looks like this and an example of use like this.
For Scala the generated code looks like this and an example of use like this.
The following basic types are supported:
type in spec | java type: req / opt |
scala type: req / opt |
---|---|---|
string |
String / String |
String / Option[String] |
int |
int / Integer |
Int / Option[Int] |
long |
long / Long |
Long / Option[Long] |
double |
double / Double |
Double / Option[Double] |
boolean |
boolean / Boolean |
Boolean / Option[Boolean] |
size |
long / Long |
Long / Option[Long] |
duration |
long / Long |
Long / Option[Long] |
duration (using --durations flag) |
Duration / Duration |
Duration / Option[Duration] |
NOTE
- please read
Optional<T>
instead of theT
values in the java "opt" column above if using the--java:optionals
flag.- using the
--durations
flag,java.time.Duration
is used instead oflong
/Long
. See durations for further information.
The size
type corresponds to the
size-in-bytes formats
supported by the Typesafe library.
See #23 for various examples.
NOTE: As of 0.0.984, a setting with a default value like
memory: 50G
(ormemory: "50G"
) will no longer be inferred as withsize
type, but just as a string (with default value"50G"
). For the size type effect, you will need to be explicit:memory: "size | 50G"
. See #42 and #41.
A duration type can be further qualified with a suffix consisting of a colon
and a desired time unit for the reported value.
For example, with the type "duration:day"
, the reported long value will be in day units,
with conversion automatically performed if the actual configuration value is given in
any other unit as supported by Typesafe Config according to the
duration format.
A more complete example with some additional explanation:
durations {
# optional duration; reported Long (Option[Long] in scala) is null (None) if value is missing
# or whatever is provided converted to days
days: "duration:day?"
# required duration; reported long (Long) is whatever is provided
# converted to hours
hours: "duration:hour"
# optional duration with default value;
# reported long (Long) is in milliseconds, either 550,000 if value is missing
# or whatever is provided converted to millis
millis: "duration:ms | 550s"
...
}
Using the --durations
flag, the reported value will be a java.time.Duration
instead of a long
/ Long
and the suffix will be ignored:
"duration:hours | 3day"
is java.time.Duration.ofDays(3)
if value is missing or whatever is provided converted to a java.time.Duration
With t denoting a handled type, a list of elements of that type
is denoted [
t ]
. The corresponding types in Java and Scala are:
type in spec | java type: req / opt |
scala type: req / opt |
---|---|---|
[ t ] |
List<T> / List<T> |
List[T] / Option[List[T]] |
where T
is the corresponding translation of t in the target language, with
List<T>
corresponding to an unmodifiable list in Java, and
List[T]
corresponding to an immutable list in Scala.
As seen in examples above, each object in the given configuration spec becomes a class.
It is of course possible to specify a field as a list of objects, for example:
positions: [
{
lat: double
lon: double
}
]
In Java this basically becomes:
public class Cfg {
public final java.util.List<Cfg.Positions$Elm> positions;
public static class Positions$Elm {
public final double lat;
public final double lon;
}
}
and in Scala:
case class Cfg(
positions : List[Cfg.Positions$Elm]
)
object Cfg {
case class Positions$Elm(
lat : Double,
lon : Double
)
}
An object or a list in the input configuration can be marked optional with
the @optional
annotation (in a comment):
#@optional
email {
server: string
password: string
}
#@optional
reals: [ { foo: double } ]
In Scala this basically becomes:
case class Cfg(
email : Option[Cfg.Email],
reals : Option[List[Cfg.Reals$Elm]]
)
object Cfg {
case class Email(
password : String,
server : String
)
case class Reals$Elm(
foo : Double
)
}
As with basic types, the meaning of an optional object or list is that the corresponding
value will be null
(or Optional.empty()
) (None
in Scala) when the corresponding actual entry is missing in
a given configuration instance.
Since version 0.9.94 we started adding support for "shared objects" (#54),
a feature that has been enhanced in later versions.
This is exercised by using the @define
annotation:
#@define
Struct {
c: string
d: int
}
example {
a: Struct
b: [ Struct ]
}
In this example, the annotation will only generate the definition of the
corresponding class Struct
in the wrapper but not the member of that
type itself. Then, the type can be referenced for other definitions.
Note: The
@define
annotation is only supported for objects and enumerations (see below), not for a basic types or lists.
As of 0.9.98 shared objects now support simple inheritance by an abstract superclass. The following syntax can be used to define a simple inheritance:
#@define abstract
BaseStruct {
a: [string]
b: double
}
#@define extends BaseStruct
ChildStruct {
c: string
d: string
}
example {
child: ChildStruct
}
In this example, the annotation will generate an abstract class definition of the BaseStruct
as well as a definition
of ChildStruct
which extends BaseStruct
. This inheritance structure simplifies processing of the config with
structs that have multiple common fields.
Only leaf members of the inheritance tree may be instantiable instances, all other shared objects in between have to be abstract classes. The following are valid definition comments:
Comment | Meaning |
---|---|
#@define abstract |
Root of an inheritance tree |
#@define abstract extends Foo |
Intermediate member of the tree, that extends the shared object Foo |
#@define extends Bla |
Leaf member of the inheritance tree, that extends the (abstract) shared object Bla |
- the current support for shared objects as field types in another shared object is unstable and not yet fully supported
- empty structs without any fields are treated as strings. Hence, having a child struct without new fields in addition to its superclass is not supported yet
Enumerations can also be defined and this is done through the
@define enum
annotation:
#@define enum
FruitType = [apple, banana, pineapple]
fruit: FruitType
someFruits: [FruitType]
other: {
aFruit: FruitType
}
As with other uses of @define
, the enumeration annotation will only generate
the enumeration type itself, but the associated name can then be used for other
field definitions in your configuration schema.
The type defined in the example above basically gets translated into Java and Scala as follows:
public enum FruitType {
apple,
banana,
pineapple;
}
sealed trait FruitType
object FruitType {
object apple extends FruitType
object banana extends FruitType
object pineapple extends FruitType
}
tscfg can also generate user-oriented configuration templates from the given configuration schema. See this wiki.
But I can just access the configuration values directly with Typesafe Config and even put them in my own classes
Sure. However, as the number of configuration properties and levels of nesting increase, the benefits of automated generation of the typesafe, immutable objects, along with the centralized verification, shall become more apparent. All of this –worth emphasizing– based on an explicit schema for the configuration.
Any tscfg best practice for my development workflow?
Please see this wiki.
Is there any sbt plugin for tscfg that I can use as part of the build for my project?
Not implemented yet. The issue is #21 if you want to add comments or reactions. PRs are also welcome.
Can tscfg generate Optional<T>
for optional fields?
Use the --java:optionals
flag for enabling Optional<T>
instead of null
for optional fields in java.
What happened with the generated toString
method?
We think it's more flexible to let client code decide how to render configuration instances while also recognizing that very likely typical serialization libraries are already being used in the application. For example, the demo programs JavaUse and scalaUse use Gson and pprint, respectively. Although you could also use Typesafe Config itself for rendering purposes, you would be using the original Typesafe Config parsed configuration object, so the rendering won't necessarily be restricted only to the elements captured in the configuration specification used by tscfg for the generated wrapper.
https://github.com/carueda/tscfg/tree/main/src/test/scala/tscfg