Struct pearl::Builder

pub struct Builder { /* private fields */ }

Is used to initialize a Storage.

Required params:

• work_dir - where Storage will keep blob and index files
• max_blob_size - upper limit of blob file size
• max_data_in_blob - maximum number of records in one blob
• blob_file_name_prefix - prefix in blob file name pattern: {prefix}.{id}.{ext}
• key_size - const key size in bytes

Example

use pearl::{Builder, Storage, ArrayKey};

let storage: Storage<ArrayKey<8>> = Builder::new()
        .blob_file_name_prefix("benchmark")
        .max_blob_size(10_000_000)
        .max_data_in_blob(1_000)
        .work_dir(std::env::temp_dir().join("pearl_benchmark"))
        .build()
        .unwrap();

Implementations

impl Builder

pub fn new() -> Self

Create new uninitialized Builder

pub fn build<K>(self) -> Result<Storage<K>>where for<'a> K: Key<'a> + 'static,

Creates Storage based on given configuration, returns error if not all params are set.

Errors

Return error if some of the required params is missed or wrong

pub fn work_dir<S: Into<PathBuf>>(self, work_dir: S) -> Self

Sets working directory. If path doesn’t exists, Storage will try to create it at initialization stage.

pub fn corrupted_dir_name(self, name: impl Into<String>) -> Self

Sets directory name for corrupted files. If path doesn’t exists, Storage will try to create it at initialization stage.

pub fn max_blob_size(self, max_blob_size: u64) -> Self

Sets blob file size approximate limit. When the file size exceeds it, active blob update is activated. Must be greater than zero

pub fn max_data_in_blob(self, max_data_in_blob: u64) -> Self

Limits max number of records in a single blob. Must be greater than zero

pub fn blob_file_name_prefix<U: Into<String>>( self, blob_file_name_prefix: U ) -> Self

Sets blob file name prefix, e.g. if prefix set to hellopearl, files will be named as hellopearl.[N].blob. Where N - index number of file If the prefix is empty, param won’t be set.

pub fn allow_duplicates(self) -> Self

Disables check existence of the record on write. Writing becomes faster because there is no additional disk access for searching for duplicates.

pub fn ignore_corrupted(self) -> Self

Ignores blobs with corrupted header for index file and continues work with normal ones (also leaves error logs for bad blobs). It’s worth noticing that index file maybe corrupted likely in case of irregular stop of the storage: either the disk disconnected or server crashed. In these cases blob file is likely either corrupted, so there is no a try to restore indices from blob file.

pub fn set_filter_config(self, config: BloomConfig) -> Self

Sets custom bloom filter config, if not set, use default values.

pub fn set_validate_data_during_index_regen(self, value: bool) -> Self

Enables or disables data checksum validation during index regeneration

pub fn create_work_dir(self, create: bool) -> Self

[Optional] Sets whether to create work directory if its missing on storage initialization. Default value is true

pub fn set_io_driver(self, iodriver: IoDriver) -> Self

Set io driver to be used during file io operations. If the driver is not set, then sync is used as the default

pub fn set_dump_sem(self, dump_sem: Arc<Semaphore>) -> Self

Sets semaphore for index dumping on blob change. Parallel saving of indexes onto the disk will be limited by this semaphore. This can prevent disk overusage in systems with multiple pearls.

pub fn set_bloom_filter_group_size(self, size: usize) -> Self

Sets bloom filter group size

pub fn set_deferred_index_dump_times(self, min: Duration, max: Duration) -> Self

Set min and max waiting time for deferred index dump

pub fn set_max_dirty_bytes_before_sync(self, bytes: u64) -> Self

Set max dirty bytes before filesync will be performed in background

Trait Implementations

impl Debug for Builder

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Builder

fn default() -> Builder

Returns the “default value” for a type.