12-Factor Apps suggest that you configure your application using environment variables. So far, so good. I can easily imagine that this is a good way to do it if you need to set a connection string, e.g.
But what if you have more complex configuration with lots and lots of values? I for sure do not want to have 50+ environment variables, do I?
How could I solve this, and still be compliant to the idea of 12-Factor Apps?
From a quick read of the configure link you provided, I agree with the author's claim that there is a widespread problem, but I am not convinced that their proposed solution is going to always be best. Like you, I don't relish the idea of having to define dozens of environment variables to configure an application. So here are some alternative ideas.
First, read Chapter 2 of the Config4* Getting Started Guide (disclaimer: I am the main author of that software). In particular, notice that its support for what I call adaptive configuration can go a long way towards addressing the concern that you ask about. Is Config4* the ultimate solution? Possibly not, but I think it is a good step in the right direction.
Second, the chances are that whatever application you are developing/maintaining has already settled on a particular configuration technology, such as XML files or Java property files, and it won't be feasible to migrate to using Config4*. This raises the question: is there anything you can do to avoid having a proliferation of, say, XML-based configuration files when you have multiple environments (such as dev, UAT, staging and production) in which the application will be deployed? I have outlined an approach for dealing with this issue in another StackOverflow article.
I saw this tutorial for writing a JSON-RPC server for SWI-Prolog. Unfortunately, all it does is add two numbers. I'm wondering if there exists a RPC server for SWI-Prolog that can define new rules and answer general Prolog queries, returning JSON lists, etc?
When you take a tour on SWI-Prolog website, proudly self-powered, you can see at work some of the features offered by http package.
It's a fairly large range of tools, and to grasp the basic of the system, the easiest way it's to follow the specific How to section, step by step. There is a small bug you should be aware in the LOD Crawler: add an option on line 42 of lod.pl:
...
; rdf_load(URI2, [format(xml)]),
....
or you will probably get
Internal server error
Domain error: content_type' expected, found text/xml;charset=UTF-8'
when running the sample.
An important feature of the IDE it's the ability to debug the HTTP requests.
When done with the HowTo, you can take a look to Cliopatria, dedicated to interfacing RDF to HTML. It come with a pirates demo, I must say I find it a bit too 'crude' for my taste, and I don't know about YUI, used in the award winning MultimediaN project. Then I've used Bootstrap to gain a modern look for the front end, with appreciable result (I'm sorry I can't - yet - publish it, need more time to engineering the system).
HTH
I am in the process of building interactive front-ends to a
distributed application which to date has been used to run workloads
that had a batch-job like structures and needed no UI at all. The application is mostly written in Perl and C and runs on a mix of Unix and Windows machines, but I think this isn't relevant to the UI.
The first such frontend is going have a command-line user interface --
currently, I envision something similar to the CLIs of the Procurve
switches and Cisco routers that I have worked with.
Like modern network gear CLIs, commands are going to resemble
simple sentences, (i.e. show vlans ports 1-4) and the CLI will
have some implicit state, much in the way that Unix shells and
cmd.exe in Windows have environment variables and current working
directories. Moreover, I'd like to implement great tab completion that
is aware of the application's state as much as possible and I want to be able to do that with as
little application-specific code as possible.
The low-level functionality (terminal I/O) seems easy to implement on
top of GNU Readline or similar libraries, but that's only where the
real fun starts. So far I have looked at the Perl modules
Term::Shell
and
Term::ShellUI,
but I'm not convinced that I want to use either of them. I am still
considering rolling my own solution and at the moment I am primarily looking for
inspiration.
Can you recommend any application or library, regardless of
implementation language, that implements a good CLI from which I can
borrow ideas?
I suggest you take a look at the philosophy underlying Microsoft PowerShell. From the idea of piping typed objects between commands to the consistency of its commands and argument syntax, I think it can be a source of inspiration.
You could try having a look at libcli :
"Libcli provides a shared library for
including a Cisco-like command-line
interface into other software."
http://code.google.com/p/libcli/
BTW - I forgot to mention that it is GNU Lesser GPL and actually used by Cisco in some products.
As for your last sentence/question, I'm particularly fond of zsh completion and line editing (zle).
How does one study open-source libraries code, particularly standard libraries?
The code base is often vast and hard to navigate. How to find some function or class definition?
Do I search through downloaded source files?
Do I need cvs/svn for that?
Maybe web-search?
Should I just know the structure of the standard library?
Is there any reference on it?
Or do some IDEs have such features? Or some other tools?
How to do it effectively without one?
What are the best practices of doing this in any open-source libraries?
Is there any convention of how are sources manipulated on Linux/Unix systems?
What are the differences for specific programming languages?
Broad presentation of the subject is highly encouraged.
I mark this 'community wiki' so everyone can rephrase and expand my awkward formulations!
Update: Probably didn't express the problem clear enough. What I want to, is to view just the source code of some specific library class or function. And the problem is mostly about work organization and usability - how do I navigate in the huge pile of sources to find the thing, maybe there are specific tools or approaches? It feels like there should've long existed some solution(s) for that.
One thing to note is that standard libraries are sometimes (often?) optimized more than is good for most production code.
Because they are widely used, they have to perform well over a wide variety of conditions, and may be full of clever tricks and special logic for corner cases.
Maybe they are not the best thing to study as a beginner.
Just a thought.
Well, I think that it's insane to just site down and read a library's code. My approach is to search whenever I come across the need to implement something by myself and then study the way that it's implemented in those libraries.
And there's also allot of projects/libraries with excellent documentation, which I find more important to read than the code. In Unix based systems you often find valuable information in the man pages.
Wow, that's a big question.
The short answer: it depends.
The long answer:
Some libraries provide documentation while others don't. Standard libraries are usually pretty well documented, whether your chosen implementation of the library includes documentation or not. For instance you may have found an implementation of the c standard library without documentation but the c standard has been around long enough that there are hundreds of good reference books available. Documentation with hyperlinks is a very useful way to learn a new API. In any case the first place I would look is the library's main website
For less well known libraries lacking documentation I find two different approaches very helpful.
First is a doc generator. Nearly every language I know of has one. It basically parses an source tree and creates documentation (usually as html or xml) which can be used to learn a library. Some use specially formatted comments in the code to create more complete documentation. JavaDoc is one good example of this. Doc generators for many other languages borrow from JavaDoc.
Second an IDE with a class browser. These act as a sort of on the fly documentation. Some display just the library's interface. Other's include description comments from the library's source.
Both of these will require access to the libraries source (which will come in handy if you intend actually use a library).
Many of these tools and techniques work equally well for closed/proprietary libraries.
The standard Java libraries' source code is available. For a beginning Java programmer these can be a great read. Especially the Collections framework is a good place to start. Take for instance the implementation of ArrayList and learn how you can implement a resizeable array in Java. Most of the source has even useful comments.
The best parts to read are probably whose purpose you can understand immediately. Start with the easy pieces and try to follow all the steps that are hidden behind that single call you make from your own code.
Something I do from time to time :
apt-get source foo
Then new C++ project (or whatever) in Eclipse and import.
=> Wow ! Browsable ! (use F3)
What are your opinions on developing for the command line first, then adding a GUI on after the fact by simply calling the command line methods?
eg.
W:\ todo AddTask "meeting with John, re: login peer review" "John's office" "2008-08-22" "14:00"
loads todo.exe and calls a function called AddTask that does some validation and throws the meeting in a database.
Eventually you add in a screen for this:
============================================================
Event: [meeting with John, re: login peer review]
Location: [John's office]
Date: [Fri. Aug. 22, 2008]
Time: [ 2:00 PM]
[Clear] [Submit]
============================================================
When you click submit, it calls the same AddTask function.
Is this considered:
a good way to code
just for the newbies
horrendous!.
Addendum:
I'm noticing a trend here for "shared library called by both the GUI and CLI executables." Is there some compelling reason why they would have to be separated, other than maybe the size of the binaries themselves?
Why not just call the same executable in different ways:
"todo /G" when you want the full-on graphical interface
"todo /I" for an interactive prompt within todo.exe (scripting, etc)
plain old "todo <function>" when you just want to do one thing and be done with it.
Addendum 2:
It was mentioned that "the way [I've] described things, you [would] need to spawn an executable every time the GUI needs to do something."
Again, this wasn't my intent. When I mentioned that the example GUI called "the same AddTask function," I didn't mean the GUI called the command line program each time. I agree that would be totally nasty. I had intended (see first addendum) that this all be held in a single executable, since it was a tiny example, but I don't think my phrasing necessarily precluded a shared library.
Also, I'd like to thank all of you for your input. This is something that keeps popping back in my mind and I appreciate the wisdom of your experience.
I would go with building a library with a command line application that links to it. Afterwards, you can create a GUI that links to the same library. Calling a command line from a GUI spawns external processes for each command and is more disruptive to the OS.
Also, with a library you can easily do unit tests for the functionality.
But even as long as your functional code is separate from your command line interpreter, then you can just re-use the source for a GUI without having the two kinds at once to perform an operation.
Put the shared functionality in a library, then write a command-line and a GUI front-end for it. That way your layer transition isn't tied to the command-line.
(Also, this way adds another security concern: shouldn't the GUI first have to make sure it's the RIGHT todo.exe that is being called?)
Joel wrote an article contrasting this ("unix-style") development to the GUI first ("Windows-style") method a few years back. He called it Biculturalism.
I think on Windows it will become normal (if it hasn't already) to wrap your logic into .NET assemblies, which you can then access from both a GUI and a PowerShell provider. That way you get the best of both worlds.
My technique for programming backend functionality first without having the need for an explicit UI (especially when the UI isn't my job yet, e.g., I'm desigining a web application that is still in the design phase) is to write unit tests.
That way I don't even need to write a console application to mock the output of my backend code -- it's all in the tests, and unlike your console app I don't have to throw the code for the tests away because they still are useful later.
I think it depends on what type of application you are developing. Designing for the command line puts you on the fast track to what Alan Cooper refers to as "Implementation Model" in The Inmates are Running the Asylum. The result is a user interface that is unintuitive and difficult to use.
37signals also advocates designing your user interface first in Getting Real. Remember, for all intents and purposes, in the majority of applications, the user interface is the program. The back end code is just there to support it.
It's probably better to start with a command line first to make sure you have the functionality correct. If your main users can't (or won't) use the command line then you can add a GUI on top of your work.
This will make your app better suited for scripting as well as limiting the amount of upfront Bikeshedding so you can get to the actual solution faster.
If you plan to keep your command-line version of your app then I don't see a problem with doing it this way - it's not time wasted. You'll still end up coding the main functionality of your app for the command-line and so you'll have a large chunk of the work done.
I don't see working this way as being a barrier to a nice UI - you've still got the time to add one and make is usable etc.
I guess this way of working would only really work if you intend for your finished app to have both command-line and GUI variants. It's easy enough to mock a UI and build your functionality into that and then beautify the UI later.
Agree with Stu: your base functionality should be in a library that is called from the command-line and GUI code. Calling the executable from the UI is unnecessary overhead at runtime.
#jcarrascal
I don't see why this has to make the GUI "bad?"
My thought would be that it would force you to think about what the "business" logic actually needs to accomplish, without worrying too much about things being pretty. Once you know what it should/can do, you can build your interface around that in whatever way makes the most sense.
Side note: Not to start a separate topic, but what is the preferred way to address answers to/comments on your questions? I considered both this, and editing the question itself.
I did exactly this on one tool I wrote, and it worked great. The end result is a scriptable tool that can also be used via a GUI.
I do agree with the sentiment that you should ensure the GUI is easy and intuitive to use, so it might be wise to even develop both at the same time... a little command line feature followed by a GUI wrapper to ensure you are doing things intuitively.
If you are true to implementing both equally, the result is an app that can be used in an automated manner, which I think is very powerful for power users.
I usually start with a class library and a separate, really crappy and basic GUI. As the Command Line involves parsing the Command Line, I feel like i'm adding a lot of unneccessary overhead.
As a Bonus, this gives an MVC-like approach, as all the "real" code is in a Class Library. Of course, at a later stage, Refactoring the library together with a real GUI into one EXE is also an option.
If you do your development right, then it should be relatively easy to switch to a GUI later on in the project. The problem is that it's kinda difficult to get it right.
Kinda depends on your goal for the program, but yeah i do this from time to time - it's quicker to code, easier to debug, and easier to write quick and dirty test cases for. And so long as i structure my code properly, i can go back and tack on a GUI later without too much work.
To those suggesting that this technique will result in horrible, unusable UIs: You're right. Writing a command-line utility is a terrible way to design a GUI. Take note, everyone out there thinking of writing a UI that isn't a CLUI - don't prototype it as a CLUI.
But, if you're writing new code that does not itself depend on a UI, then go for it.
A better approach might be to develop the logic as a lib with a well defined API and, at the dev stage, no interface (or a hard coded interface) then you can wright the CLI or GUI later
I would not do this for a couple of reasons.
Design:
A GUI and a CLI are two different interfaces used to access an underlying implementation. They are generally used for different purposes (GUI is for a live user, CLI is usually accessed by scripting) and can often have different requirements. Coupling the two together is not a wise choice and is bound to cause you trouble down the road.
Performance:
The way you've described things, you need to spawn an executable every time the GUI needs to d o something. This is just plain ugly.
The right way to do this is to put the implementation in a library that's called by both the CLI and the GUI.
John Gruber had a good post about the concept of adding a GUI to a program not designed for one: Ronco Spray-On Usability
Summary: It doesn't work. If usability isn't designed into an application from the beginning, adding it later is more work than anyone is willing to do.
#Maudite
The command-line app will check params up front and the GUI won't - but they'll still be checking the same params and inputting them into some generic worker functions.
Still the same goal. I don't see the command-line version affecting the quality of the GUI one.
Do a program that you expose as a web-service. then do the gui and command line to call the same web service. This approach also allows you to make a web-gui, and also to provide the functionality as SaaS to extranet partners, and/or to better secure the business logic.
This also allows your program to more easily participate in a SOA environement.
For the web-service, don't go overboard. do yaml or xml-rpc. Keep it simple.
In addition to what Stu said, having a shared library will allow you to use it from web applications as well. Or even from an IDE plugin.
There are several reasons why doing it this way is not a good idea. A lot of them have been mentioned, so I'll just stick with one specific point.
Command-line tools are usually not interactive at all, while GUI's are. This is a fundamental difference. This is for example painful for long-running tasks.
Your command-line tool will at best print out some kind of progress information - newlines, a textual progress bar, a bunch of output, ... Any kind of error it can only output to the console.
Now you want to slap a GUI on top of that, what do you do ? Parse the output of your long-running command line tool ? Scan for WARNING and ERROR in that output to throw up a dialog box ?
At best, most UI's built this way throw up a pulsating busy bar for as long as the command runs, then show you a success or failure dialog when the command exits. Sadly, this is how a lot of UNIX GUI programs are thrown together, making it a terrible user experience.
Most repliers here are correct in saying that you should probably abstract the actual functionality of your program into a library, then write a command-line interface and the GUI at the same time for it. All your business logic should be in your library, and either UI (yes, a command line is a UI) should only do whatever is necessary to interface between your business logic and your UI.
A command line is too poor a UI to make sure you develop your library good enough for GUI use later. You should start with both from the get-go, or start with the GUI programming. It's easy to add a command line interface to a library developed for a GUI, but it's a lot harder the other way around, precisely because of all the interactive features the GUI will need (reporting, progress, error dialogs, i18n, ...)
Command line tools generate less events then GUI apps and usually check all the params before starting. This will limit your gui because for a gui, it could make more sense to ask for the params as your program works or afterwards.
If you don't care about the GUI then don't worry about it. If the end result will be a gui, make the gui first, then do the command line version. Or you could work on both at the same time.
--Massive edit--
After spending some time on my current project, I feel as though I have come full circle from my previous answer. I think it is better to do the command line first and then wrap a gui on it. If you need to, I think you can make a great gui afterwards. By doing the command line first, you get all of the arguments down first so there is no surprises (until the requirements change) when you are doing the UI/UX.
That is exactly one of my most important realizations about coding and I wish more people would take such approach.
Just one minor clarification: The GUI should not be a wrapper around the command line. Instead one should be able to drive the core of the program from either a GUI or a command line. At least at the beginning and just basic operations.
When is this a great idea?
When you want to make sure that your domain implementation is independent of the GUI framework. You want to code around the framework not into the framework
When is this a bad idea?
When you are sure your framework will never die