GUI Framework with all Dependencies Embedded
"Raudrohi" stands for Achillea millefolium, i.e. yarrow, in Estonian. "Raud" stands for "iron" in Estonian. "Rohi" stands for "grass" and "medical drug" in Estonian.
The GitHub account (https://github.com/martinvahi/raudrohi ) contains only a crippled, incomplete, version of the RJSL. A complete development package (size: ~1GiB) can be downloaded from:
An almost complete development version can be downloaded from the http://npmjs.org by executing
npm install raudrohi
This version exists mainly for 3 reasons:
fulfill dependencies of other projects;
to allow other, a bit bold, people to experiment with the RJSL;
The documentation of it is in a pretty shoddy state. The shoddiest parts of the documentation are not even published.
The code style is dynamic language oriented: any hack that is considered to work reliably on targeted web browsers, is applied, regardless of how many "best practices" and "good code style" rules it violates.
The output of various code style policing software, JSLint, JSHint, etc., is intentionally ignored. Despite the very hackish nature of the code, almost every design decision has been evaluated from speed and memory usage point of view.
The RJSL can have a high learning curve and the architecture of the RJSL has been optimized to minimize the amount of work that software developers, who have crossed the learning curve, have to do to implement a web application. Nothing is, and hopefully never will be, dumbed down, regardless of the barrier of entry that the solution imposes. In Albert Einstein style: the RJSL is kept as simple as possible, but not simpler.
Backwards compatibility between different versions is, intentionally, NOT MAINTAINED.
The RJSL consists of global namespaces, which wrap functions that can be used by ignoring all the rest of the RJSL.
target instance specific ID as a string, i.e. "instance phone number"
origin instance specific ID as a string
message type as a string, i.e. target instance can use that value to reroute the message to an instance specific method that processes that kind of messages.
To send a message to the web server, one sends a message to a specific widget instance that acts as a gateway to the server. Server responses are received by the gateways, which then repackage the response to the message objects and send the message objects to the widget that is targeted by the response.
The properties of the widgets probably depend on the RJSL version. The API is always subject to refactoring and no backwards compatibility is maintained.
The RJSL framework allows the creation of composite widgets by using HTML and designating the positions of subwidgets by DIV-tags. The custom widgets can be reused for composing new widgets and the custom widgets fit into the RJSL framework as if they were builtin widgets.
Widgets can be non-graphical and have a role of a general building component.
Each of the widgets has an on-off state to regulate the usage of network traffic. For example, there's no point of asking data from the server before a log-in session has been established and widgets that depend on private data can be switched on and off according to the existence of a login-session.
Each of the widgets has a built-in state machine that interacts with the built-in state machines of its subwidgets. The state machine has one default, mandatory, state called "zero". Every time the "zero" state is entered, all of the widget's subwidgets are also set to the state "zero". States can be grouped into user-defined clusters and the state "zero" is by default part of every cluster. State transitions and cluster-transitions can trigger user-defined actions.
The main benefit of the widgets built-in state machine system is that it facilitates the writing of GUI business logic. That's actually, what the state machine system has been designed for, but it's possible to ignore its existence, i.e. it's not mandatory to use it.
As of October 2012 widgets' built-in state system allows user-defined states to belong to more than one cluster, but that's fundamentally flawed approach, because that way it's not possible to determine the execution order of state-cluster transition event-hanlers. If client code is written with a limitation that every user-defined state belongs at most to only one state cluster, then the refactoring of the state system will probably not break the client code.
The RJSL contains a global message passing system, that allows any RJSL widget instance to send messages to any other RJSL widget instance.
AJAX communication with the web server is normalized out by wrapping the gate to the web server into one of the non-graphical widgets.
As of October 2012 the message passing system API has to be rewritten, because one has cleaner specification for it, but the system that exists, is not fundamentally flawed. Its API and protocol are the ones that need to be heavily refactored.
The RJSL message passing system contains means for handling AJAX responses that come in too late, have become "irrelevant", or are duplicates.
Implementation overview: Every widget has a microsession counter. If a widget A sends out a message to widget B and the microsession counter of widget A is incremented before the answer from the widget B arrives to the widget A, then the answer from the widget B is dismissed, because all answer-carrying messages that have a different origin microsession counter value than the current microsession counter value are dismissed. The widget B may, but does not have to, represent the web server.
Example scenario: widget A orders a list of names from a server and displays the text "Loading..." to the user, while waiting for the answer. The user changes its mind and sets the widget A to a different state that has to display something else that the widget A has to order from the server. Widget A orders the new type of data, but receives the previously ordered list of names before the set of the new type of data. As the state of the widget A has changed, the microsession counter value of widget A has changed, the list of names is dismissed and the widget A keeps on waiting for the newer set of data.
Each of the RJSL widgets has its own microsession counter, but its not mandatory to use it, neither is it mandatory to initialize it in client code.
Historical note: the microsession architecture has emerged purely form practice and is a result of tedious bugfixing/refactoring.
The widgets visibility is interpreted in relation to the widget's parent widget. If a parent widget is visible, then only the subwidgets that have its visibility bit set, are visible. The visibility is meant to change during runtime and all of the widgets are responsible for maintaining their own data. For example, if a textarea widget is set to be hidden, then it saves its text and renders it next time, when it becomes visible again.
All graphical widgets have a "readonly mode" and "editable mode". The mode is imposed recursively. For example, if the editability bit of a report widget is set to "false", then all of its data entry fields can switch to readonly mode. That functionality allows the same, document/application specific, widgets to be used for both, displaying and editing documents.
An illustration: A text field widget changes from "text field" to "plain text", if it is switched from editable mode to readonly mode. The button widget implements the readonly/editable modes by being enabled/disabled.
Browser normalization is based on a bottom layer which wraps third party libraries.
That means that the RJSL can probably keep up with the browser evolution by swapping the third party libraries/library versions in the bottom layer, without any need for any changes in the rest of the RJSL.
Historical note: One of the first versions of the RJSL did not wrap the third party library, YUI 2.something, to a bottom layer and when it came to the YUI library update, heavy refactoring had to be done. The reason for the update was that the YUI 2.something became obsolete, did not support latest browser versions.
The Microsoft Internet Explorer is not supported and probably will never be supported.
The API of the RJSL is never even meant to be stable. It will be refactored to any extent that one feels comfortable with.
The RJSL will probably never be popular, because it is targeted to technical audience and the needs of web developers and novice/hobby programmers are intentionally ignored.
All of the RJSL, including the 3. party parts, is under a license that allows redistribution, modification and commercial, closed-source, use.
The parts that have been written by me, firstname.lastname@example.org, are placed to namespace raudrohi and are under the BSD license, except code examples in the
, which are in public domain.
Namespace liilia contains code that is not written by me, Martin Vahi, but has been modified by me. The word "liilia" stands for "lilium" in Estonian.
, which also contains all of the RJSL dependencies.
The "Hello World" resides at
and it works "out of the box".
The RJSL uses HTML5.
Projects that depend on the RJSL probably depend on an environment variable called RAUDROHI_HOME, which is meant to point to the folder that contains the README.md that You are currently reading.
Build scripts are all Linux specific. To build RJSL, the PATH must contain
Java 7, Ruby 2.x.x, Rake
The rest of the tools are bundled with the RJSL development deliverables.
Release version of the RJSL is built by
cd $RAUDROHI_HOME/src/dev_tools rake build
Debug version of the RJSL is built by
cd $RAUDROHI_HOME/src/dev_tools rake b
RJSL depends on a specific version of the mmmv_devel_tools (https://github.com/martinvahi/mmmv_devel_tools ). To save RJSL developers from the work of finding and installing the right version of the mmmv_devel_tools, the RJSL build scripts use copy of the mmmv_devel_tools from
The mmmv_devel_tools depends on a specific version of the Kibuvits Ruby Library (hereafter KRL, http://kibuvits.rubyforge.org/ ). To save the users of the mmmv_devel_tools from the work of finding and installing the right version of the KRL, the mmmv_devel_tools uses a copy of KRL that is bundled with it. The RJSL build scripts, mainly the
also uses the KRL. As of 2013_11_04 no more than one version of the KRL can be used in a single rubyscript. Consequently the mmmv_devel_tools determines the KRL version for the RJSL build scripts.
which gets the value of the Ruby constant KIBUVITS_HOME from the copy of the mmmv_devel_tools.
The first thing to refactor is the widget internal state machine related state cluster implementation that allows a user-defined state to belong to more than one cluster. A correct version is that user-defined states can belong to at most one state cluster, because then there will be no need to determine the execution order of cluster change event handler functions.
The second at the list is the messaging system related protocol. It's not necessarily flawed, but it's so terrible that it's hard to work with. One should also implement the concept of "bus-packets" or "packet-bus'es".