Structuring (physical) source and (virtual) solution folders for portability

Copying here those comments of mine at a discussion on the GraphX project:

https://github.com/panthernet/GraphX/issues/21

describing the source code (physical) folder structure and the Visual Studio solution (virtual) folder structure I’ve been using at ClipFlair and other multi-platform projects.

——

looking at the folders/projects/libraries/namespaces naming, I think it would be more appropriate to add the platform at the end of the name, say GraphX.somePart.PCL, GraphX.somePart.UWA (=Universal Windows Application model [aka Win10]) etc. Not sure how easy it will be though for contributors to merge pending changes via GitHub if you do such drastic changes (I’m still struggling with Git myself, prefer Mercurial)

Speaking of moving the platform to the end of the name (e.g. GraphX.Controls.WPF, GraphX.Controls.SL5 etc.), to do it on the folder names (apart from doing at the source code for packages, target assemblies etc.), I think one has to use some Version Control command to "move" (Git should have something like that) the files to the new folder, else other contributors may have issue merging their changes.

Despite the trouble to do it, I think it is better for the long run. Also, all GraphX.SomePart.* subfolders could then be grouped in a GraphX.SomePart folder as subfolders where a Source or Common subfolder would also exists that has all the common code those platform-specific versions of GraphX.SomePart share (via linked files to ..\Common\SomeFile, ..\Common\SomeFolder\SomeFile and ..\Common\SomeOtherFolder\SomeFile etc.)

This is the scheme I’ve been using (and I’m very satisfied with) at http://clipflair.codeplex.com and other projects (e.g. at the AmnesiaOfWho game [http://facebook.com/AmnesiaOfWho] that has separate versions for SL5, WP7 and WP8 and WPF version on the works, plus WRT (WinRT) and UWA [Universal Windows App] too coming in the near future, with all code and most XAML shared via linked files and UserControl[s])

I meant I’d expect GraphX.Common folder with GraphX.Common.PCL in it and GraphX.Common.WP7 etc. versions for example also in that folder since PCL is usually set at WP8 level. This is just an example.

What I’m saying is that the platform name is the last part of the specialization chain, so logically GraphX comes first then say comes Controls then comes WPF, SL5, WP8 etc. in the folder name.

Also would have 3 parent folders Controls, Common and Logic. The last two would just contain the respective .PCL subfolder for now, but I can contribute subfolders for specific platforms too, esp for those the common PCL profile doesn’t cover. Source would be at Controls\Source, Common\Source, Logic\Source and respective platform specific projects (even the pcl projects) would use linked files

aka

GraphX (contains GraphX.sln)

–\Controls
—-\Source
—-\GraphX.Controls.WP7
—-\GraphX.Controls.WP8
—-\GraphX.Controls.SL5 (contains GraphX.Controls.SL5.csproj)
—-\GraphX.Controls.WPF (contains GraphX.Controls.WPF.csproj)
—-\GraphX.Controls.WIN8
—-\GraphX.Controls.UWA
—- … (more platform specific versions)

–\Common
—-\Source
—-\GraphX.Common.PCL
—- … (platform specific versions, esp. those not covered by the settings chosen at the PCL)

–\Logic
—-\Source
—-\GraphX.Logic.PCL
—- … (platform specific versions, esp. those not covered by the settings chosen at the PCL)

The same structure would also be used at examples to cover the potential of porting some of them to more platforms:

–\Examples
—-\SomeExample
——\Source
——\SomeExample.WPF
——\SomeExample.SL5
—-\SomeOtherExample
——\Source
——\SomeOtherExample.WIN8
——\SomeOtherExample.UWA
etc.

Of course common code at each folder is at the Source subfolder of that folder, shared using linked files (e.g. at \Examples\SomeExample\Source) and platform specific code is inside the respective projects (e.g. at \Examples\SomeExample\SomeExample.WPF)

Similarly the GraphX.sln would contain solution folders "Controls", "Common" and "Examples", though it could also contain separate solution (virtual) folders per platform that have each one of them "Controls", "Common" and "Examples" in them. That is the solution’s folder structure is organized per-platform in such a case. This is mostly useful if you want to focus on a specific platform each time when developing. However since the solution folders are virtual, one could even go as far as having two solutions, one with the same structure as the filesystem folders I suggest and one with a per-platform/target structure.

I follow the per-platform virtual solution folders style at ClipFlair.sln in http://clipflair.codeplex.com, while the real folders are structured as I describe above (where each module has its own physical subfolders for the various platforms). In fact some module subfolders there (say Client\ZUI\ColorChooser) contain their own extra solution file when I want to be able to focus just on a certain module. That solution just includes the respective ColorChooser.WPF, ColorChooser.SL5 etc. subprojects from respective subfolders). Such solutions also contain virtual WPF, Silverlight etc. subfolders that has as children apart from the respective platform-specific project (say ColorChooser.WPF) any other platform-specific projects needed (e.g. ….\Helpers\Utils\Utils.WPF\Utils.WPF.csproj etc.) by that module to compile.

Speaking of Xamarin, adding support for that too could follow the same pattern as described above (PCL where possible and platform-specific versions for .XamarinIOS, .XamarinAndroid etc. where needed)

Source code analyzers for .NET porting & Portable Class Libraries (PCL)

HowTo: Use MEF to implement import/export etc. plugin architecture

Copying here my comment at a discussion on the GraphX project:

https://github.com/panthernet/GraphX/pull/15

in case it helps somebody in using MEF (Managed Extensibility Framework) in their software’s architecture

——–

Using static classes instead of interfaces can mean though that you need to use reflection to call them (e.g. if you wan to have a list of export plugins).

Instead can keep interfaces and make use of MEF to locate import/export and other plugins (you can have some class attribute there that mark the class as a GraphXExporter and MEF can be asked then to give you interface instances from classes that have that attribute.)

see usage at
http://clipflair.codeplex.com/SourceControl/latest#Client/ClipFlair.Windows/ClipFlair.Windows.Map/MapWindowFactory.cs

//Project: ClipFlair (http://ClipFlair.codeplex.com)
//Filename: MapWindowFactory.cs
//Version: 20140318

using System.ComponentModel.Composition;

namespace ClipFlair.Windows.Map
{

  //Supported views
  [Export("ClipFlair.Windows.Views.MapView", typeof(IWindowFactory))]
  //MEF creation policy
  [PartCreationPolicy(CreationPolicy.Shared)]
  public class MapWindowFactory : IWindowFactory
  {
    public BaseWindow CreateWindow()
    {
      return new MapWindow();
    }
  }

}

and at
http://clipflair.codeplex.com/SourceControl/latest#Client/ClipFlair.Windows/ClipFlair.Windows.Image/ImageWindowFactory.cs

//Project: ClipFlair (http://ClipFlair.codeplex.com)
//Filename: ImageWindowFactory.cs
//Version: 20140616

using System.ComponentModel.Composition;
using System.IO;

namespace ClipFlair.Windows.Image
{

  //Supported file extensions
  [Export(".PNG", typeof(IFileWindowFactory))]
  [Export(".JPG", typeof(IFileWindowFactory))]
  //Supported views
  [Export("ClipFlair.Windows.Views.ImageView", typeof(IWindowFactory))]
  //MEF creation Policy
  [PartCreationPolicy(CreationPolicy.Shared)]
  public class ImageWindowFactory : IFileWindowFactory
  {

    public const string LOAD_FILTER = "Image files (*.png, *.jpg)|*.png;*.jpg";

    public static string[] SUPPORTED_FILE_EXTENSIONS = new string[] { ".PNG", ".JPG" };

    public string[] SupportedFileExtensions()
    {
      return SUPPORTED_FILE_EXTENSIONS;
    }

    public BaseWindow CreateWindow()
    {
      return new ImageWindow();
    }

  }

}

then at
http://clipflair.codeplex.com/SourceControl/latest#Client/ClipFlair.Windows/ClipFlair.Windows.Base/Source/View/BaseWindow.xaml.cs

to get the first plugin that supports some contract (I get that contract name from the serialization file [using DataContracts]) for a loaded view, or that supports some file extension for a file dropped inside a component, I do:

    protected static IWindowFactory GetWindowFactory(string contract)
    {
      Lazy<IWindowFactory> win = mefContainer.GetExports<IWindowFactory>(contract).FirstOrDefault();
      if (win == null)
        throw new Exception(BaseWindowStrings.msgUnknownViewType + contract);
      else
        return win.Value;
    }

    protected static IFileWindowFactory GetFileWindowFactory(string contract)
    {
      Lazy<IFileWindowFactory> win = mefContainer.GetExports<IFileWindowFactory>(contract).FirstOrDefault();
      if (win != null)
        return win.Value;
      else
        return null;
    }

Fix: Enable Silverlight & other NPAPI plugins at Chrome web browser

Below I’m elaborating a bit more my related tweet:

 

Showing below the easiest of the suggested solutions that I found in this page

 

At Chrome’s address bar you type:

chrome://flags/#enable-npapi

and press the ENTER key on the keyboard

image

Then you should see a page Chrome generates to change some of its internal settings. When NPAPI is disabled the respective entry should appear in grey background like below.

image

Press Enable at the setting “Enable NPAPI Mac, Windows”

After enabling NPAPI the page should look like this (with the respective setting in white background):

image

After enabling the NPAPI option, close the Chrome webbrowser and reopen it.

 

You can then test if Silverlight is working by visiting for example

http://studio.clipflair.net (ClipFlair Studio Foreign language learning application) or http://zoomicon.com/amnesiaofwho (Amnesia of Who memory game)

 

Btw, Chrome also is available as a Windows 8 app, in which mode it probably doesn’t support plugins at all, so if you’re running it on Windows 8 and see it always full screen inside a scrolling container, use the Chrome menu from the top-right of its window and select the option there to switch to the desktop version of Chrome instead (should say "Relaunch Chrome in desktop mode")

HowTo: Open page from Internet Explorer (Metro) app into desktop IE

The Windows 8/8.1 app version of Internet Explorer is also known as IE Metro because of the “Metro” codename (inspired by navigation signs in public transport] of the Modern UI design language promoted by Microsoft).

However that version isn’t the full Internet Explorer, in that it is unfortunately not supporting extensibility via plugins in the form of ActiveX controls as the classic (desktop version) of IE. It is only embedding the Flash player engine directly in its codebase, but not Microsoft’s own Rich Internet Application (RIA) rendering engine aka Silverlight, nor Unity or other VRML/X3D, QuickTime/QuickTimeVR etc. plugins.

Browser pages cannot detect the difference between running IE on the desktop or as an app, there is however a workarround for webpage authors or webadmins to force the app version of IE to show a prompt to the user that allows the opening of a page in the desktop version of Internet Explorer. There is also a way for System Administrators to set specific sites to open in the desktop version of IE without the user seeing such prompt.

At https://msdn.microsoft.com/en-us/library/ie/hh968248.aspx, Microsoft mentions:

As a web developer, you can enable the requiresActiveX feature switch either by using this HTTP header:

X-UA-Compatible: requiresActiveX=true

Or by using this meta element on each affected webpage:

<meta http-equiv="X-UA-Compatible" content="requiresActiveX=true"/>

 

I just added the meta tag inside the <head>…</head> block of the Amnesia of Who web version that uses Silverlight and here is how it shows in the IE Metro version (note that Silverlight IS installed in that Windows 8.1 machine, it’s just that it’s not available in that browser, that’s why the Silverlight installation prompt is also shown):

image

When the user presses the default button “Open on the desktop”, the OS switches to classic desktop mode and shows an Internet Explorer window with the Silverlight application starting fine (or if Silverlight is not installed it will prompt and allow the user to install it – note that Silverlight ActiveX control’s installation doesn’t need administrator permissions since that installation doesn’t affect other users, nor requires any elevated rights in the system to work).

image

 

I hope that Microsoft, apart from keeping on supporting this workarround, will do a clever move this time and embed Silverlight too (apart from the Flash engine that was in IE Metro) in the Spartan browser that it prepares as the Windows 10 default touch browser. And why not, provide some extensibility method for it, since HTML5 cannot become a huge, impossible to implement beast, that covers every future conceived functionality for the web.

HowTo: Delete all nodes and relationships from Neo4j graph database

At a Neo4j question in http://stackoverflow.com/questions/19624414/delete-node-and-relationships-using-cypher-query-over-rest-api, a recent reply (older ones use obsolete Cypher syntax) says:

Both START and the [r?] syntax are being phased out. It’s also usually not advised to directly use internal ids. Try something like:

match (n{some_field:"some_val"})
optional match (n)-[r]-()
delete n,r

So, to delete all nodes (including disconnected ones) and their relationships you could do:
MATCH (n) OPTIONAL MATCH (n)-[r]-() DELETE n,r
(nice that it works in a single line too)

However since we delete ALL nodes and relationships, this one looks cleaner:
MATCH (n), ()-[r]-() DELETE n,r

 

Update #1

Michael Hunger kindly commented below on this last query:

In your last query you create a huge cross product.
All nodes times all relationships.

Probably cleaner then to split it into two, delete rels first then nodes

Indeed, using PROFILE before the query it seems to do more work from a quick look (hoped it would be a bit more clever to optimize this, but maybe I’m asking too much). So probably should change it to two queries:

MATCH ()-[r]-() DELETE r

and

MATCH (n) DELETE n

 

Update #2

For deleting really big graphs, checkout an answer by Stefan Armbruster on how to delete in an iterative way at

http://stackoverflow.com/questions/29711757/best-way-to-delete-all-nodes-and-relationships-in-cypher/29715865

…the most easy way is to stop Neo4j, drop the data/graph.db folder and restart it.

Deleting a large graph via Cypher will be always slower but still doable if you use a proper transaction size to prevent memory issues (remember transaction are built up in memory first before they get committed). Typically 50-100k atomic operations is a good idea. You can add a limit to your deletion statement to control tx sizes and report back how many nodes have been deleted. Rerun this statement until a value of 0 is returned back:

MATCH (n)
OPTIONAL MATCH (n)-[r]-()
WITH n,r LIMIT 50000
DELETE n,r
RETURN count(n) as deletedNodesCount

Gotcha: MarkerReached event of MediaElement returns new Markers

I just checked in the implementation code for a new feature for ClipFlair Studio’s Captions/Revoicing component:

When playing back recorded (or loaded from a WAV or MP3 file) audio for a caption/revoicing entry, the playback is now limited to the duration of the respective caption, (End-Start) time that is (btw that component has a duration column too that is hidden by default and can be shown by flipping it with the gear button on its titlebar and selecting the respective option to show the column).

The original audio is not affected and is stored in whole inside the saved state of the component/activity, so that you can adjust the caption entries timerange at any time to fit all or part of that recorded audio entry if you wish.

Will see into adding an “Limit playback” option to the backpanel of that component (it will default to true/checked) for any ClipFlair activities that don’t use the Start/End/Duration columns (e.g. if some activity just wants a grid of Captions and Audio entries for practicing and maybe for comparing to audio samples provided by the activity author or teacher at the optional “Comments (Audio)” column).

While implementing this feature, there were some “gotchas” that caused me some headache to spot:

1) When MediaOpened event is called by the MediaElement control, the Markers collection has just been reset and you need at that point to add your TimelineMediaMarker that will notify you when the playback limit point has been reached to stop the playback. In the case above this event is called after recording some audio or loading some WAV or MP3 audio file at an AudioRecorderControl (one is used at each row of the captions grid)

2) One shouldn’t remove an added marker at MediaEnded or MediaFailed events. This is since those will fire at each revoicing entry playback, whereas the MediaOpened while only occur once when the Audio property is populated at the AudioRecorderControl. As I mention above, MediaElement clears the Markers collection every time new content is loaded to it, so we need not worry about removing the marker we had added before.

3) Maybe the least obvious issue and the one that caused me most of headache to spot was that the MediaElement’s MarkerReached event gets back a MediaMarker that isn’t the same Marker instance as the one you had added to the Markers collection. So you have to use the Text property of the marker when you create it and then compare with the text from the one you got in the event to see if they are equal strings (btw, when C# compares strings it does it by content even if you use == instead of Equals method, unlike Java, where you shouldn’t use == to compare strings)

Follow

Get every new post delivered to your Inbox.

Join 992 other followers

%d bloggers like this: