b207464fe6 Update README.md with bdk-ffi-bindgen info (Steve Myers)
fca5d1602b Add workspace and move bin to bdk-ffi-bindgen package (Steve Myers)
f4e097c4ac Only print python fix up lib path if used (Steve Myers)
c66dfdd52a Use structopt to capture generate options (Steve Myers)
ce848725b4 Add binary to remove the need for uniffi-bindgen cli tool (thunderbiscuit)
Pull request description:
This PR is based on the mozilla/application-services [embedded-uniffi-bindgen](https://github.com/mozilla/application-services/tree/main/tools/embedded-uniffi-bindgen) tool. The purpose is to keep the bdk-ffi and bdk-ffi-bindgen tool in sync with the same version of uniffi-rs.
Fixes#124, this PR replaces #122.
The `bdkffi` library code remains unchanged but the `bin/generate` and `bin/generate-bindings` bins are combined and put in a new workspace binary package called `bdk-ffi-bindgen`. The `bdk-ffi-bindgen` binary uses the following options, defaults, and environment variables:
```shell
% cargo run -p bdk-ffi-bindgen -- --help
bdk-ffi-bindgen 0.1.0
A tool to generate bdk-ffi language bindings
USAGE:
bdk-ffi-bindgen [OPTIONS] --language <language> --out-dir <out-dir>
FLAGS:
-h, --help Prints help information
-V, --version Prints version information
OPTIONS:
-l, --language <language>
Language to generate bindings for [env: BDKFFI_BINDGEN_LANGUAGE=] [possible values: kotlin, swift, python]
-o, --out-dir <out-dir>
Output directory to put generated language bindings [env: BDKFFI_BINDGEN_OUTPUT_DIR=]
-p, --python-fixup-path <python-fixup-path> Python fix up lib path [env: BDKFFI_BINDGEN_PYTHON_FIXUP_PATH=]
-u, --udl-file <udl-file> UDL file [env: BDKFFI_BINDGEN_UDL=] [default: src/bdk.udl]
```
Top commit has no ACKs.
Tree-SHA512: fa1a1c097fe5d0e704d76078c10f82c466dad5d045c8c93d579c2d13c448c52fb6a4f99dfd3dbc46be30471477ae2d1f9264201e14bae7948b408c8e0b3c9b81
851f61296a Fix Wallet.broadcast function, now returns a tx id as a hex string (Steve Myers)
Pull request description:
ACKs for top commit:
thunderbiscuit:
Tested ACK 851f612.
Tree-SHA512: 86e1d39029924e4fa3a0c21e9f45c1ba0694f4db9d1cfd8dee25a5675d5a8b7851a7c712ce57fb74382a7428fcecbe8ecee4b7b87b9245672bbd6ccea63dfc13
cc3736809a Fix memory database configuration enum (thunderbiscuit)
Pull request description:
The `DatabaseConfig.Memory` enum currently requires a "junk" string argument which is not used when creating the wallet:
```rust
// lib.rs line 24
pub enum DatabaseConfig {
Memory { junk: String },
Sled { config: SledDbConfiguration },
}
// lib.rs line 209
impl Wallet {
fn new(
descriptor: String,
change_descriptor: Option<String>,
network: Network,
database_config: DatabaseConfig,
blockchain_config: BlockchainConfig,
) -> Result<Self, BdkError> {
let any_database_config = match database_config {
DatabaseConfig::Memory { .. } => AnyDatabaseConfig::Memory(()),
DatabaseConfig::Sled { config } => AnyDatabaseConfig::Sled(config),
};
```
Which translates to the udl file like this:
```txt
[Enum]
interface DatabaseConfig {
Memory(string junk);
Sled(SledDbConfiguration config);
};
```
According to the [docs from uniffi-rs](https://mozilla.github.io/uniffi-rs/udl/enumerations.html) the `interface` here is required because the enums have named fields. But after testing I found that we can declare the udl file like so, and remove the requirement for the `junk` argument:
```txt
[Enum]
interface DatabaseConfig {
Memory();
Sled(SledDbConfiguration config);
};
```
On the Rust side we then have
```rust
pub enum DatabaseConfig {
Memory,
Sled { config: SledDbConfiguration },
}
```
And the resulting bindings go from (note that the bindings transform the enum into a sealed class rather than a Kotlin enum)
```kotlin
sealed class DatabaseConfig {
data class Memory(
val junk: String
) : DatabaseConfig()
data class Sled(
val config: SledDbConfiguration
) : DatabaseConfig()
```
to
```kotlin
sealed class DatabaseConfig {
object Memory : DatabaseConfig()
data class Sled(
val config: SledDbConfiguration
) : DatabaseConfig()
```
Which makes the API simpler to use, and removes the confusion created by having to provide an empty string (or not know what we're supposed to provide) to the `Memory()` enum.
The final call-site looks like this:
```kotlin
fun onlineWalletSyncGetBalance() {
// val db = DatabaseConfig.Memory("")
val db = DatabaseConfig.Memory
val client = BlockchainConfig.Electrum(
ElectrumConfig(
"ssl://electrum.blockstream.info:60002",
null,
5u,
null,
100u
)
)
val wallet = Wallet(descriptor, null, Network.REGTEST, databaseConfig, blockchainConfig)
wallet.sync(LogProgress(), null)
val balance = wallet.getBalance()
assertTrue(balance > 0u)
}
```
All tests run well on my side of things, but I'm opening this more as a discussion piece because I wasn't sure if there were other reasons for the choice of providing the argument to the `Memory` enum, or other design choices I'm not aware of. Any thoughts on this @artfuldev? I think you were the one who wrote the initial enum.
Top commit has no ACKs.
Tree-SHA512: 135e5943039a08522773f721a7cf6bbb93bd5bb9394bf42a30bab5f3e16fd35ce078056756e020a666d4f574d74080bc3404cc81809c0d7e0afe5c9471878425
