Git Product home page Git Product logo

archivist's Introduction

Archivist

Archivist is a flexible data storage service for WoW AddOns. It is designed especially for addons which need to store a large amount of data, but only occasionally read or update this data. Data given to the archivist is stored in SavedVariables in a compressed format, to minimize addon load time.

Table of Contents

Using Archivist

Using the archivist is very simple. If you just wish to use the global archive, then install Archivist as a standalone addon. Note that if you use the global archive, it is recommended to set Archivist as a dependency in your .toc file.

Getting Started

First, load your data store from the archive:

-- Load data from the archive. If the RawData store MyAddonArchive doesn't exist in the archive, then it is created automatically. Usually you will use this to obtain your store.
myStore = Archivist:Load("RawData", "MyAddonArchive")

-- if you have good reason to believe that your store is archived, then you can use open:
myStore = Archivist:Open("RawData", "MyAddonArchive") -- throws an error if "MyAddonArchive" doesn't exist in the archive

-- if you know that your store doesn't yet exist, then you can call Create:
myStore = Archivist:Create("RawData", "MyAddonArchive")

-- If you wish to create an "anonymous" store for some reason, that is also supported. Archivist will auto-generate a random storeID for you. But don't lose the storeID, or it will be hard to find this data again later.
myStore, storeID = Archivist:Create("RawData")

All of the basic Archivist Verbs (Create, Load, Open, Close, Commit) take as parameters the Store Type and Store ID. The Store Type must be a string, and identifies the kind of store object you wish to create. The store type must be registered via RegisterStoreType (see Create New Store Types below for more information). The Store ID must be a string, and is unique for that given Store Type. You may have as many stores named MyAddonArchive as you like, so long as all of them have differing Store Types.

The RawData store type is just a table. You may mutate this table in any way you see fit. Once you are done with your reads and writes, you may close myStore:

Archivist:CloseStore(myStore)
-- or...
Archivist:Close("RawData", "MyAddonArchive")

Once closed, any changes to myStore will not be archived unless you reopen the store.

Some store types (see Create New Store Types below) may operate in such a way that it makes sense to commit changes to the archive, without closing the store. Archivist supports this operation too:

Archivist:CommitStore(myStore)
-- or...
Archivist:Commit("RawData", "MyAddonArchive") -- In the case of RawData this is not very useful, since its contents are also committed when the store is closed

All open stores are automatically closed and archived on PLAYER_LOGOUT. You should not try to read or write data from a store on or after that event. The exact behavior at that point depends on the implementation of the Store Type, as well as the order in which WoW dispatches events to addon scripts, and is outside the Archivist specification.

Included Store Types

Archivist comes prepackaged with some basic store types, both for your convenience and as an example for implementing your own. They are listed here:

  • RawData
    • A simple table with no extra bells or whistles. The contents of this table are stored directly into the archive when this archive is committed or closed.
  • Readonly
    • Identical to RawData, except that Commit & Close methods don't do anything. This causes Archivist to never overwrite data in the savedvariables for these stores, once created.

Embedding the Archivist

In most use cases, you will want to embed Archivist, so that your addon's archive does not intersect with that of any other addon. Embedding the Archivist into your addon is very similar, but a few more steps are needed. First, create an archive addon:

## Interface: 90005
## Title: MyArchive
## LoadOnDemand: 1
## SavedVariables: MyArchiveSaved

You may also use your addon's native SavedVariables file if you wish (instead of creating a secondary addon), but this will cause your addon to always load its archive from the disk, which may not be desirable.

Archivist follows the library standard (though it is not a library) for making itself available to be embedded. To include Archivist in your addon, it is pretty easy if you already use other libraries.

First, include Archivist as an external in your .pkgmeta file:

externals:
  MyAddon/Embeds/Archivist: https://github.com/emptyrivers/Archivist

Then, in your embeds.xml (if you use one), you can include Archivist.xml:

<Ui xmlns="http://www.blizzard.com/wow/ui/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.blizzard.com/wow/ui/ ..\FrameXML\UI.xsd">
	<!-- Other embedded addons...-->
	<Include file=".\Embeds\Archivist.xml"/>
</Ui>

That's it! Archivist is now available in your addon releases.

You can also include Archivist via your .toc file instead, if you prefer.

To use Archivist in your addon code:

local addon, ns = ...

-- Note that Archivist embeds itself in your addon's namespace.
local Archivist = ns.Archivist

-- when you have need to access the archive, then load the data...
LoadAddOn("MyArchive")
-- and have Archivist initialize your archive.
local archive = Archivist(MyArchiveSaved)

Now, you may use the Archivist just like you would if it had been installed in a standalone form.

Managing Multiple Archives

Starting with version 2.0, Archivist supports management of multiple savedvariables with a single embed.

NOTE: this is a breaking change if you are upgrading from a tag earlier than v2.0.0. Specifically, the following change is needed:

local Archivist = select(2, ...).Archivist

Archivist:Initialize(MyArchiveSaved) -- Bad! the Archivist is no longer itself an archive
local store = Archivist:Load(...) -- attempt to call method "Load" (a nil value)

local archive = Archivist:Initialize(MyArchiveSaved) -- Good!
local store = archive:Load("RawData", "MyStore")

local otherArchive = Archivist(MyOtherArchiveSaved) -- for your convenience, Archivist has a __call metamethod that points to Initialize

local otherStore = otherArchive:Load("RawData", "MyStore")

assert(store ~= otherStore) -- since they were loaded from different archives, these stores will always be independent from one another

Custom Store Types

Some use cases demand more sophisticated data management. Archivist can accomodate this need, with custom Store Types:

local prototpye = {
	id = "MyStoreType",
	version = 1,
	Init = function() end, 
	Create = function(...) end,
	Open = function(image) end,
	Update = function(image) end,
	Commit = function(store) end,
	Close = function(store) end,
	Delete = function(image) end,
	Wind = function(image) end,
	Unwind = function(woundImage) end,
}

Archivist:RegisterStoreType(prototype)

Prototype Fields are as follows:

  • id
    • Unique Identifier of the store type, e.g. RawData.
  • version
    • Version number of store type. Useful if the prototype changes in a backwards incompatible way, and archived data needs to be massaged before use.

Prototype methods:

  • Init
    • Optional function. Initialize your prototype. If provided, then Init is always guaranteed to run exactly once per game session, before any other method is run.
  • Create
    • Create a new, empty, store object. Extra arguments passed into Archivist:Create will be passed into this function if you wish to accomodate initial setup of the store object.
    • Must return non-nil store object.
  • Open
    • Create from data an active store object.
    • Must return non-nil store object
  • Update
    • Optional function. If provided, then archived data is replaced with the return value of Update. If no change is needed, then return nil.
  • Commit
    • Return image of data to be archived
  • Wind/Unwind
    • Optional functions. If provided, these are used instead of Archive/DeArchive to transfer data to/from a storable/openable state.
    • It is an error to provide one, but not both, of Wind/Unwind.
    • Useful for when a mixed archive is desirable (e.g. so that the savedvariables are legible without spinning up Archivist for debug/support reasons)
  • Close
    • Deactivate store. Returned value will be written to archive. If no update to archive is needed, then return nil.
    • Once close is called on a store, Archivist will not update the archived data again unless the store is opened.
  • Delete
    • Called after store is closed & no longer exists in SV. Useful for deleting any dependent stores.

Store Type Methods Should be Functional

With the exception of Init, all of these functions may be called at any time without warning. Thus, they should ideally be written as close to purely functional as possible, with few-to-no side effects. If you must have side effects, then do your best to write functions whose side effects are idempotent. This will help you avoid weird problems from functions being called in an order you didn't expect, and other hard-to-debug behaviors.

Re-registering prototypes

If multiple independent codebases share an archive, and they both register the same store type, then what happens depends on the version number:

  • If the second registration has equal or lower version number, then the second registration is ignored.
  • If the second registration has a higher version number, then:
    • Each active store is Closed using the old Close method
    • Each Archived store is Updated using the new Update method if provided
    • The new Init method is run.
    • Any previously open stores are opened.

Re-registering a store type is generally not recommended, as you risk data loss and other errors. If you are using a store type in a shared archive which you expect to be registered multiple times (e.g. by independent custom code auras in WeakAuras), then it is recommended to only keep stores of that type open for the time that you need them to be open. Alternatively, design the Initialization routine such that it can broadcast the re-initialization, so that any system that was holding a reference to any open store can re-obtain that reference.

Performance

Archivist stores data in a compressed format, to minimize load times. This means that to open an archive, Archivist must fully decompress the data. Depending on the level of compression, and the size before compression, this can potentially take a long time to do.

To circumvent this, consider breaking your data into chunks (ideally into chunks that are meaningful for your code, and not arbitrarily based on data size), and archiving each piece in its own store, if you believe that your archive will contain a large amount of data. The Check verb is provided as a cheap way to check if a given storeID exists in the archive.

Full API List

-- Main API

-- Opens (or creates) the given store. This is the main entry point for your addon's code.
store = archive:Load(storeType, storeID)
-- or... (though rarely useful - if storeID is nil then Load is an alias for Create)
store, storeID = archive:Load(storeType)

-- Register store type. Store type must be registered before an archive can be accessed. All verbs will raise an error if called with an unregistered storeType.
archive:RegisterStoreType(prototype)


-- Archivist verbs
-- 	These are the main "actions" that archivist knows how to do.

-- Creates a new archive, and returns an active store object. Raises an error if archive already exists. In most cases you'll want to use Load instead.
store = archive:Create(storeType, storeID, ...)
-- or... (though rarely useful)
store, storeID = archive:Create(storeType)

-- Opens an archive, and returns active store object. Raises an error if archive doesn't yet exist. In most cases you'll want to use Load instead.
store = archive:Open(storeType, storeID)

-- Commit to archive without closing store. This will cause a change in the archive.
-- 	Store is still considered open after committing it.
archive:CommitStore(store)
-- or...
archive:Commit(storeType, storeID)

-- Closes a store. Once closed, you may discard the store object.
-- 	May also update archive, depending on the store type. Once closed, you should consider the store object obsolete.
-- 	Occurs automatically on PLAYER_LOGOUT.
archive:CloseStore(store)
-- or...
archive:Close(storeType, storeID)
-- Note: once a store is closed, manipulating the old store object is considered outside of the Archivist specification, and behavior depends on the store type implementation. However, any changes after Close will *never* be archived

-- Check if a store exists.
-- 	This verb is intended for performance-critical operations, where you only need to ensure that the store exists. Guaranteed to never call Archive/DeArchive, or invoke any prototype methods.
-- Returns true if data for the given storeID exists in the archive, or an active store object exists.
-- Otherwise, returns false.
storeExists = archive:Check(storeType, storeID)

-- Close and delete store permanently. Use ONLY if you are absolutely sure you don't want the data anymore. Archivist cannot help you retrieve lost data once you invoke this.
-- 	If force is truthy, then the Delete will go through even if the store type is not registered. This is useful if you decide to drop support for a storeType.
archive:DeleteStore(store, force)
-- or...
archive:Delete(storeType, storeID, force)
-- or... (if you want to delete everything)
-- If storeType is given, then all stores of that type are deleted. If storeType is not given, then all stores in the entire archive are deleted. This is provided to assist in cases like, "the user wishes to destroy all data and start fresh".
-- USE WITH CAUTION. YOUR DATA WILL BE LOST. YOU WILL NOT GET A SECOND CHANCE.
archive:DeleteAll(storetype)

-- Create identical but independent copy of archive. If openstore is truthy, then also Opens the cloned archive and returns active store object.
store = archive:CloneStore(store, openStore)
-- or...
store = archive:Clone(storeType, storeID, openStore)


-- Plumbing Methods
-- 	The following methods are used internally, and are usually not useful for addons using the Archivist.

-- transforms savedvariables data into image suitable for Open()
image = archive:Unwind(storeType, woundImage)

-- transforms image suitable for Open() into data to be stored in savedvariables
woundImage = archive:Wind(storetype, image)

-- Generate a random uuid. Used when Create is called without providing a storeID
uuid = archive:GenerateID()

-- Compress data. Data passed in is not touched in any way, and calling code may retain ownership.
compressedString = archive:Archive(data)

-- Decompress data. String is expected to have been compressed using Archivist:Archive.
-- 	Compressing, and then decompressing the same data is essentially an expensive clone operation.
data = archive:DeArchive(compressedString)

-- Close all stores immediately. Automatically called on PLAYER_LOGOUT. Not usually useful for addons using 
archive:CloseAllStores()

Limitations and "Gotchas"

Like every other project, Archivist can't do everything. The major restrictions (and consequences of these limitations) are described here.

Archivist Modifies Your Addon's Namespace

Specifically, when Archivist is embedded into your addon, then the Archivist field on your namespace table (the one you access with local addon, ns = ...) is set to the Archivist object. I don't expect many people have ever used that particular field on their namespace table, but please do be aware of this if you decide to use the archivist.

For this reason, you also shouldn't embed Archivist in a library intended for 3rd party consumption via LibStub. Anybody who embeds your addon may not be expecting their namespace to be modified, and unexpected changes to the internal namespace is pretty rude to force on someone.

Archivist Can't Store Everything

Archivist is a WoW addon, and it cannot do anything that the WoW addon environment does not allow for. In practice, this means that Archivist cannot serialize functions or closures it receives, as there is no way to convert these to a string format. If your store type has functions which need to be stored (e.g. generated code), then instead of trying to archive the function directly, store all of the information which was used to generate the function, and in your Open procedure, re-generate the function based on the data image you receive.

Additionally, Archivist does not check for metatables. So, if your store type uses __index or __newindex metamethods in any way, then providing the store object when asked for a data image will likely result in data loss. To avoid this, ensure that your Close and Commit methods return objects that contain all of the data you need to archive directly, without needing to invoke any metamethods.

Archivist is Not a Library

In the WoW addon world, we are used to embedded code that is "library-style". That is, everyone shares the same code, and shares the same versioning. This is where tools like LibStub and the Ace framework come into play. These are great when you just want to use code that somebody else already wrote, without worrying too much about implementation details. Archivist even uses this for compression, because I don't want to have to reinvent the wheel just to compress some data. Also, LibDeflate is really good.

Archivist explicitly does not follow this pattern. In a slightly different world, Archivist might have been designed to fit the LibStub paradigm, but the specifics of the WoW AddOn environment (and what Archivist is supposed to do) make that the wrong pattern to follow, in my opinion. Instead, Archivist is a service you may embed into your code, but it does not share its code or internal memory with other instances of Archivist in other addons. Don't rely on some other commonly installed addon to register a store type you need for you. That's bad practice in any scenario, and it just won't work with Archivist.

Please don't register Archivist with LibStub. This will "work", in that you won't get any errors. But if somebody else does the same thing, then one of you will be unable to access your own archive, since Archivist is designed to be run and initialized by each addon independently for each instance. If a user installs two addons which both use Archivist like this, then nothing will seem amiss for a while until one of these addons is disabled and suddenly the other "loses" all of its data. Then you get angry users, and nobody likes angry users.

You can set Archivist as a hard dependency in your .toc file, if you want to use a "global" archive. In this way, Archivist can behave nicely while still being "shared code". However, a global archive is like writing code with lots of global variables - not such a great idea. The standalone Archivist addon is distributed primarily so that it can be embedded into other addons easily via curseforge or packaging scripts, as well as for the use of individual users in private, non-shared setups that just need a place to store some data without too much hassle.

archivist's People

Contributors

emptyrivers avatar lwafaure avatar meorawr avatar stanzilla avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar  avatar

archivist's Issues

DeleteAll with store type breaks Archivist for rest of session

self.sv[storeType] = nil

The current implementation of the Archivist:DeleteAll(storeType) function nils out the named store table in self.sv when a store type is explicitly passed. This leads to the rest of Archivist breaking if you then call other functions for the store type that was just deleted.

If called without an explicit store type the self.sv entries for each store type aren't nilled out and are instead replaced with empty tables, leaving Archivist functioning fine when other functions are called.

Very basic test case:

Archivist:Initialize({});
print(Archivist.sv.ReadOnly);
Archivist:DeleteAll("ReadOnly");
print(Archivist.sv.ReadOnly);
Archivist:Create("ReadOnly", "TestStore", ":)");  -- Error! :(
print(Archivist.sv.ReadOnly.TestStore);

This will throw the following error once it calls the Create function:

4x SecretProject\Libs\Archivist\Archivist.lua:200: attempt to index field '?' (a nil value)
[string "@SecretProject\Libs\Archivist\Archivist.lua"]:200: in function `Create'

Compared to the alternative of not passing a store name, no error occurs:

Archivist:Initialize({});
print(Archivist.sv.ReadOnly);
Archivist:DeleteAll();
print(Archivist.sv.ReadOnly);
Archivist:Create("ReadOnly", "TestStore", ":)");
print(Archivist.sv.ReadOnly.TestStore);

Error with /wa repair

Using your branch from WeakAuras/WeakAuras2#2067
When typing /wa repair i get this error

1x WeakAuras\Libs\Archivist\Archivist.lua:555: Encountered unexpected token type while parsing object. Expected "KEY" but got "VALUE".
[string "=[C]"]: in function `error'
[string "@WeakAuras\Libs\Archivist\Archivist.lua"]:53: in function `Assert'
[string "@WeakAuras\Libs\Archivist\Archivist.lua"]:555: in function <WeakAuras\Libs\Archivist\Archivist.lua:533>
[string "@WeakAuras\Libs\Archivist\Archivist.lua"]:574: in function `DeArchive'
[string "@WeakAuras\Libs\Archivist\Archivist.lua"]:299: in function <WeakAuras\Libs\Archivist\Archivist.lua:290>
[string "=(tail call)"]: ?
[string "=(tail call)"]: ?
[string "@WeakAuras\ArchiveTypes\Repository.lua"]:56: in function <WeakAuras\ArchiveTypes\Repository.lua:52>
[string "=(tail call)"]: ?
[string "=(tail call)"]: ?
[string "@WeakAuras\WeakAuras-@[email protected]"]:2824: in function <WeakAuras\WeakAuras.lua:2818>

SavedVariables.zip

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.