fmap

A simple and efficient approach to caching or persistent blob storage. Map-like object for accessing data stored on the filesystem.

var fmap = Fmap(directory);

fmap['keyA'] = 'my string';         // saved string into a file
fmap['keyB'] = 777;                 // saved int into a file
fmap['keyC'] = [0x12, 0x34, 0x56];  // saved three-bytes into a file

print(fmap['keyA']); // read from file

Fmap implements a Map, so it can be used the same way.

print('Count of items: ${fmap.length}');

for (var entry in fmap.entries) {
    print('Item ${entry.key}: ${entry.value}'); 
}

Entries are stored in separate files. Therefore, reading/writing an entry is about as fast as reading/writing a file with a known exact name. But unlike direct file access, Fmap has no restrictions on the content of String keys, it takes care of finding unique file names, resolving hash collisions, and serialization.

Creating

For persistent data storage

var fmap = Fmap(Directory('/path/to/my_precious_data'));

To cache data in the system temporary directory

var fmap = Fmap.temp(); // will be placed into {temp}/fmap dir

To cache data in a specific subdirectory of the system temporary directory

var images = Fmap.temp(subdir: 'images_cache'); // {temp}/images_cache
var jsons  = Fmap.temp(subdir: 'jsons_cache');  // {temp}/jsons_cache

If all the storage items have the same type, you can set the type with generics

var strings1 = Fmap<String>(directory);
var strings2 = Fmap.temp<String>();

Types

The collection allows you to store only values of certain types. Supported types are String, List<int>, List<String>, int, double and bool.

var fmap = Fmap(directory);
fmap['string'] = 'abcd';
fmap['int'] = 5;
fmap['double'] = 5.0; 
fmap['bool'] = true;

Bytes

Any List<int> is treated as list of bytes.

fmap['blob1'] = [0x12, 0x34, 0x56]; // List<int>
fmap['blob2'] = utf8.encode('my string'); // List<int>
fmap['blob3'] = myFile.readAsBytesSync(); // Uint8List implements List<int>

Since numbers are bytes, each int inside a list is truncated to the range 0..255.

fmap['blob4'] = [1, 10, -1, 777]; // saves 1, 10, 255, 9

Lists of bytes always return Uint8List when read.

Uint8List bytes = fmap['blob4']; // was List<int>, now Uint8List

String lists

Lists of strings, when stored, can be specified by any object that implements Iterable <String>, not necessarily a List.

fmap['saved_list'] = ['ordered', 'strings', 'in', 'list'];
fmap['saved_iterable'] = {'unordered', 'strings', 'in', 'a', 'set'};

However, when read, they will definitely return as List<String>.

List<String> a = fmap['saved_list'];
List<String> b = fmap['saved_iterable'];

Entry ~ file

Keep in mind that each entry is saved in a separate file. Therefore, storing a lot of atomic values like double associated with different keys may not be very practical. Conversely, saving large objects such as String, List<int>, or List<String> is efficient. It's almost like writing directly to files, but without restrictions on key names.

Purging

If the storage has become too large, you can delete the oldest data.

// leave only the newest 16 megabytes
fmap.purge(16*1024*1024);

Which elements are removed depends on the policy argument passed to the constructor.

var fmap = Fmap(dir, policy: Policy.fifo);

Two policies are supported: FIFO and LRU. By default, this is FIFO.

If you want the purge method to purge storage with LRU policy, you must not only create Fmap(policy: Policy.lru) before purging but always create the object this way. It will force Fmap to update the last-used timestamps every time an entry is read.

When you do not specify this argument, the timestamps are only updated on writes, but not on reads.

Compatibility

The library is unit-tested on Linux, Windows, and macOS. Mobile systems such as Android and iOS have the same kernels as their desktop relatives.

Libraries

fmap