lmdb-cluster
A clustered version of lmdb that supports an HTTP API with sockets planned.
LMDB functionality for put, get, remove, and range queries with versions is supported.
Also supported are:
- basic patch operations for partial object updates via lmdb-patch,
- selective patch operations through extended paths,
- copy operations using the HTTP verb
COPY - move operations using the HTTP verb
MOVE - a query mechanism that supports functional, declarative and RegExp filters via lmdb-query.
- a search mechanism based on indexes via lmdb-index.
LMDB start-up options for encryption, compression, etc. are also supported.
This is ALPHA software. Full unit tests not yet in place. API may change. Not functionally complete.
See also:
LMDB Index - Higher level object operations for LMDB values, e.g. indexing and indexed based queries.
LMDB Query - A higher level query mechanism for LMDB supporting functional, declarative and RegExp filters without the overhead of an entire database wrapper.
LMDB IndexedDB - An IndexedDB wrapper for LMDB that supports the full IndexedDB API.
Installation
npm install lmdb-cluster
Usage
See file server.js in test directory and unit test cases in index.test.js for examples until documentation complete.
Also see the LMDB documentation.
Starting Server
const serve = require("lmdb-cluster").serve;
// Start a server with an environment called "test" that uses versioing for all databases.
// Include a database called "test" that uses default lmdb options for the database.
serve({environments:{test:{useVersions:true,databases:{test:null}}}}).then((server) => {
console.log("Server started");
});HTTP API
General API approach is to use the HTTP verbs to indicate the database function to be called:
- COPY - no underlying LMDB operation, but the capability is provided by
lmdb-copy. - DELETE - removeXXX
- GET - getXXX, plus querying based on URL query string parameters when a key is not provided
- MOVE - no underlying LMDB operation, but the capability is provided by
lmdb-move. - PATCH - no underlying LMDB operation, but the capability is provided by
lmdb-patch. - PUT - putXXX
- POST - transaction and other activities that are not simple database operations. (not yet implemented)
The last portion of a URL path is typically a URLEncoded string that is the key for the database operation, although when the key is missing, a range get, query, search or indexed object action is presumed.
A query string is used to pass arguments to the database operation, with the exception of data to be put into the database, which is passed in the body of a request.
We start with serve because nothing else will work without it! After that, we will look at the API in alphabetical order.
serve(options:object|string):Promise
Starts a server. Returns a promise that resolves when the server is ready.
If options is an object, it has the surface:
{
serverOptions = {},
clusterOptions = {},
appOptions = {},
databaseOptions = {}
}If the options argument to serve is a string, it is the path to a JavaScript file that exports the options in the above form. The path is resolved relative to the current working directory, i.e. the directory from which the file containing the call to serve is saved. See the 'test' directory for an example.
serverOptions are the same options that would be passed to NodeJS server.listen.
clusterOptions are the same options that would be used for NodeJS clustering, this includes maxCPUs and on event handlers, e.g.
{
maxCPUs: 2,
on: {
online(worker) {
console.log(`Worker ${worker.process.pid} is online`);
}
}
}The this context of event functions will be bound to the cluster. If you do not provide an on.exit handler, one will be provided for you to restart the worker.
exit(worker, code, signal) {
console.log(`Worker ${worker.process.pid} died with code: ${code}, and signal: ${signal}`);
console.log('Starting a new worker');
cluster.fork();
}appOptions are the same options that would be passed to the Fastify factory function.
databaseOptions has the surface and defaults below:
{
dbroute="/data/:environment/:name",
defaultEnvironment = {}, // was environmentOptions = {},
dynamicEnvironment, // wasdynamicEnvironmentOptions,
dynamicDatabase={} // was dynamicDatabaseOptions=dynamicEnvironmentOptions,
environments = {},
}You can change the dbroute if you wish to use a different API endpoint. For example, /dbs/:environment/:name would be a valid value. You must include :environment and :name in the route.
defaultEnvironment, dynamicEnvironment and dynamicDatabase are all optional and have the surface {options:object,functions:object}. The options are the same that would be passed to the LMDB open or openDB. If defaultEnvironment is present, then it is used as the default environment for all databases. If dynamicEnvironment is present, then it is used as the default for any environment not configured in the environments argument. It can also have inheritDefaults as a boolean property. If dynamicDatabase is present, then it is used for any database not configured in the environments argument. It can also have inheritEnvironment as a boolean property.
environments is an object that has the surface:
{
<environmentName>: {
options: {},
functions: {},
inheritDefaults: false, // merges defaultEnvironment.option and defaultEnvironment.functions with environment options and functions
databases: {
<databaseName>: {
inheritEnvironment: false, // merges environment.options and environment.functions with database options and functions
options: null,
functions: null,
}
}
}
}If dynamicEnvironment is present and an attempt is made to open an environment not configured in the environments argument, then the dynamicEnvironment.options are used to create an environment database of the name requested. Otherwise, an error is thrown. Be careful with this, as it can be a DOS security risk.
dynamicDatabaseOptions are used in the same way as dynamicEnvironmentOptions for child databases. The dynamicEnvironmentOptions are automatically used if neither the parent environment is configured to have it's options inherited or dynamicDatabaseOptions is present. Be careful with this, as it can be a DOS security risk.
The return values below are JSON encoded in the body of the response.
COPY /data/:environment/:name/:key?key=:newKey,&version=:version&ifVersion=:ifversion - returns boolean
Copies the value at :key to :newKey. The newKey will be in the same :environment and database. If :version is provided, it will be used for the copy. Copying will only occur when the version of the original matches :ifversion, when :ifversion is provided. Returns true if successful, otherwise false.
DELETE /data/:environment/:name/:key?ifVersion=:ifversion - returns boolean
Deletes a value from the database :name with :key. Optionally, only deletes if the version of the value is :ifversion.
Returns true or false depending on whether the value was deleted, i.e. if the value did not exist or the version did not match false is returned.
GET /data/:environment/:name/:key?ifVersion=:ifversion&entry=:entry - returns any value
Gets a value from the database :name with :key. If :entry is true, then the value is returned as an object with the properties value and version.
Returns the value or null if the value does not exist or the version does not match :ifVersion when provided.
GET /data/:environment/:name/:key/*?ifVersion=:ifversion - returns any value
Gets a value from the database :name with :key and subpath *. Optionally, only gets if the version of the root value is :ifversion.
Returns the value or null if the root or leaf value does not exist or the version does not match.
GET /data/:environment/:name/?start=:start&end=:end&indexMatch=:indexMatch&keyMatch=:keyMatch&valueMatch=:valueMatch&select=:select&limit=:limit&offset=:offset&reverse=:reverse&versions=:versions - returns result object
Gets a range of values from the database :name. Note: Trailing slash is REQUIRED.
All query string arguments are optional. However, failing to provide one of :start, :indexMatch or keyMatch will result in a full database scan.
Only one of :start, :indexMatch or keyMatch can be provided. And, :end is only used with :start.
The :start and :end are JSON.stringified URLEncoded strings. The :start and end follow the same form as LMDB range options.
Alternatively, :keyMatch, can be used to specify a range for getRangeWhere from lmdb-query. Like :start and :end, this is a JSON.stringified URLEncoded string.
Or, :indexMatch uses a JSON.stringified URLEncoded string with lmdb-index to search the database and returns a range of values.
The remaining parameters limit=:limit&offset=:offset&reverse=:reverse&versions=:versions&version="version are the same.
The :limit is the maximum number of values to return.
The offset is the number of values to skip before returning values. It is used in conjunction with :limit to implement paging.
The :reverse is a boolean that indicates whether the range should be returned in reverse order (not yet implemented)
The :versions is a boolean that indicates whether the version of each value should be returned.
And :version is used to return a value only if the version matches.
Result Object
Returns a result object with the surface that is similar to the standard iterable result object:
{
done: <boolean> or absent,
value: <array>
offset: <number>
}The items in the value array have the surface:
{
key: <key>,
value: <value>,
version: <version> // optionally
}If offset is present, then the range is not complete. Request the exact same URL as before, but with the offset set to the value of the offset property of the result object. This will return the next page of results. When a JavaScript client library is written, the object will even have a next() method.
MOVE /data/:environment/:name/:key?key=:newKey,&version=:version&ifVersion=:ifversion - returns boolean
Moves the value at :key to :newKey. The newKey will be in the same :environment and database. If :version is provided, it will be used for the moved version. Moving will only occur when the version of the original matches :ifversion, when :ifversion is provided. Returns true if successful, otherwise false. The actual move involves making a copy and deleting the original inside a transaction.
PATCH /data/:environment/:name/:key?version=:version&ifVersion=:ifversion - returns boolean
Patches a value (the contents of the request body) into the database :name at the :key, using Object.assign, optionally with the version provided. If ifVersion is provided, only patches if the version of the old value is :ifversion.
If the existing value is a primitive type or the new value is a primitive type, the existing value is replaced with the patch. Otherwise, the patch is merged into the object. By serializing undefined to "@undefined" as the value for properties you wish to delete, PATCH can be used to delete properties (including nested ones).
PATCH /data/:environment/:name/:key/*?version=:version&ifVersion=:ifversion - returns boolean
Patches a value (the contents of the request body) into the database :name at the :key, and subpath *, optionally with the version provided. If ifVersion is provided, only patches if the version of the old value is :ifversion.
This approach is useful when you wish to patch portions of a sub-object without needing a copy of the parent data. For example, if you have a document with a user object, you can patch the user.address.zip without needing a copy of the data outside of address or even the city and state in the address sub-object. This might not only more efficient, but also more secure.
If the existing value is a primitive type or the new value is a primitive type, the existing value is replaced with the patch. Otherwise, the patch is merged into the object. By serializing undefined to "@undefined" as the value for properties you wish to delete, PATCH can be used to delete properties (including nested ones).
PUT /data/:environment/:name/?version=:version&ifVersion=:ifversion - returns id
Puts an object value (note no key is provided above), assigns an id if needed, indexes the object, and returns the id. The object is the contents of the request body. If :version is provided, it will be used for the version. Putting will only occur when the version of the original matches :ifversion, when :ifversion is provided.
Currently, all top level properties on all objects are indexed. Nested properties are not indexed. This may change in the future.
PUT /data/:environment/:name/:key?version=:version&ifVersion=:ifversion - returns boolean
Puts a value (the contents of the request body) into the database :name at the :key, optionally with the version provided. If ifVersion is provided, only puts if the version of the old value is :ifversion. Note, if you put a value with a :key indexing is NOT conducted. If you wish to index the value, use PUT /data/:environment/:name/?version=:version&ifVersion=:ifversion instead.
FAQS
- Why doesn't the library use ESM modules?
The Socket.io library and some of the other libraries used by this library are more "friendly" to Common JS and it is easy to pull ESM modules into Common JS modules. It is not as easy to pull Common JS modules into ESM modules. So, for now, this library is Common JS only.
2, What is the primary underlying tech?
Socket.io, Fastify, and of course LMDB from Symas.
Roadmap
v2.0.0 - Socket API
Release History (Reverse Chronological Order)
2023-04-23 v0.6.1 Updated lmdb-query and lmdb-index to better support index based queries.
2023-04-23 v0.6.0 Added support for index based queries.
2023-04-22 v0.5.0 Added support for PUT of objects without a key plus indexing.
2023-04-20 v0.4.0 Now using lmdb-copy, lmdb-move, and lmdb-extend for COPY, MOVE instead of local code. Added support for targeted nested get and patching.
2023-04-19 v0.3.0 Added support for COPY and MOVE.
2023-04-18 v0.2.0 Enhanced documentation. Added support for keyMatch, valueMatch, and select for GET methods through the use of lmdb-query.
2023-04-15 v0.1.0 Enhanced documentation. Added support for PATCH. BREAKING CHANGE, the signature of the databaseOptions argument to serve` has changed.
2023-04-14 v0.0.5 Enhanced documentation.
2023-04-14 v0.0.4 Enhanced documentation. Implemented loading options from disk.
2023-04-13 v0.0.3 Refined arguments to serve. Enhanced documentation.
2023-04-12 v0.0.2 Added support for HTTPS.
2023-04-11 v0.0.1 Initial public release