23

I am working on a first PCL that targets : WSA (Windows Store Application), WPF,WP7,WP8. We can say that it is a rolerdex kind of application, you have contacts , they have contact details and images. (it's not, but I can't give details about the application, so I am using a very simple example instead). Here are some of my questions :)

  1. Should I have the images in the PCL?

If yes:

  1. How do reference the image for usage in WSA?
  2. How do I best solve scaling with the scale qualifiers etc. when used across different projects?

I am not using a database and the images are not downloaded from an external service- I would like to keep the images (not many really) locally, in the app or in the PCL.

EDIT: I just want to display images. That's it. It's a static rolerdex, you can't add new people. I just want to display 5 number of people and their image (in the PCL). How do I reference the images if it's a Windows Store Application?

I have a binding and the DataContext is set to a ViewModel in the PCL. The ViewModel aggregates the data to be displayed from the models. The property I've bound against is MyImage. Ignoring the other platforms, how would the Uri look like? Everything else works fine.

I really just want help with these three questions, although I really appreciate all the answers!!!

Iris Classon
  • 5,752
  • 3
  • 33
  • 52
  • Its just a poor attempt att shortning the darn long name, Windows Store Application. I've made sure it says so in the Q now :) – Iris Classon Sep 20 '12 at 12:38
  • I think the problem might be the word "reference" - if you mean Uri? Then take a look at something like http://stackoverflow.com/questions/10511946/get-uri-for-a-stored-file-in-storagefile-winrt ? – Stuart Sep 20 '12 at 13:22
  • That Uri scheme does not work with the PCL, but maybe something similar? At the same time I honestly don't know if it is possible. I've googled for a day now, and tried dozen variations. I was hoping somebody knew for sure :/ – Iris Classon Sep 20 '12 at 13:28
  • From my answer... in "the PCL control code will be able to treat each file as just a pair of strings - the folder it's in and the file name" - together they make an "address" for your image - then afterwards use "a specific ValueConverter to generate the correct filename or filestream on each platform" – Stuart Sep 20 '12 at 13:34
  • My question still remains the same :/ , how would the Uri look like? I'm feeling pretty dumb at this point when somebody explains to me how a path looks like (or am I misunderstanding your comment?) – Iris Classon Sep 20 '12 at 14:01
  • You could call it the Iris Uri - "iris://folder/filename" - obviously converters would need to map it to "isostorage", to "ms-appdata", etc? – Stuart Sep 20 '12 at 14:14
  • Have you been able to make that work with a PCL and WinRT app? I've tried quite a few versions all day with no luck, but maybe I've missed something. I would love a a working Uri example (a line of code, not a whole project) WinRT => Image in PCL . Relative paths are not supported in WinRT, so I'm curious what else would be a problem.. – Iris Classon Sep 20 '12 at 14:22
  • I'm not sure what you're asking now. Maybe try a new clean question about winrt - give the xaml and c# you have working in wp7 or wpf - and ask a specific question about how to translate that to winrt. Get it working in rt first, then add PCL. its new territory for all of us, but we can get there 1 step at a time. – Stuart Sep 20 '12 at 17:51

7 Answers7

23

For a lot of cases, images are platform-specific. They need to cater for size and DPI of the device itself, and would need to fit in with the look and feel of the application. For these situations, I would have the View itself decide what images to show to the user, probably based on some sort of state/mode provided by the ViewModel.

However, these are cases where the images need to come from the ViewModel, for example, in the case of the sender thumbnails that get displayed in mail applications. In these cases, I have the ViewModel return some sort of a platform-agnostic concept of an image (such as byte[]), and then have the platform-specific projects convert that into something that their UI stack understands (in XAML, this would be a ImageSource).

The code would look something like this:

Portable project:

using System.IO;
using System.Reflection;

namespace Portable
{
    public class ViewModel
    {
        private byte[] _image = LoadFromResource("Image.png");

        public byte[] Image
        {
            get { return _image; }
        }

        private static byte[] LoadFromResource(string name)
        {
            using (Stream stream = typeof(ViewModel).GetTypeInfo().Assembly.GetManifestResourceStream("Portable." + name))
            {
                MemoryStream buffer = new MemoryStream();
                stream.CopyTo(buffer);

                return buffer.ToArray();
            }
        }
    }
}

Note: You will need to remove or add GetTypeInfo() depending on the platforms you are targeting.

Here we're reading from an embedded resource (Properties -> Build Action -> Embedded Resource), but you could imagine this coming from the network, or somewhere else.

Windows Store app project: In the Windows Store app, you would have a value converter to convert from byte[] -> ImageSource:

using System;
using System.IO;
using Windows.Storage.Streams;
using Windows.UI.Xaml.Data;
using Windows.UI.Xaml.Media.Imaging;

namespace App
{
    public class ByteToImageSourceValueConverter : IValueConverter
    {
        public object Convert(object value, Type targetType, object parameter, string language)
        {
            InMemoryRandomAccessStream s = new InMemoryRandomAccessStream();

            byte[] bytes = (byte[])value;
            Stream stream = s.AsStreamForWrite();
            stream.Write(bytes, 0, bytes.Length);
            stream.Flush();
            stream.Seek(0, SeekOrigin.Begin);


            BitmapImage source = new BitmapImage();
            source.SetSource(s);

            return source;           

        }

        public object ConvertBack(object value, Type targetType, object parameter, string language)
        {
            throw new NotImplementedException();
        }
    }
}

In the code behind of the View, set the DataContext:

DataContext = new ViewModel();

Then in the View itself binding to the ViewModel.Image property, and set the converter:

<Page.Resources>
    <local:ByteToImageSourceValueConverter x:Name="ImageConverter"/>
</Page.Resources>

<Grid >
    <Image HorizontalAlignment="Left" Height="242" Margin="77,10,0,0" VerticalAlignment="Top" Width="278" Source="{Binding Image, Converter={StaticResource ImageConverter}}"/>

</Grid>
David Kean
  • 5,722
  • 26
  • 26
  • 1
    It's worth pointing out that this is a performance killer for most applications where there are several images and the images have significant size. It also undermines the ability for the underlying framework to scale the images to different devices. I believe a better solution might be to first cache the images and reference their paths. – Jerry Nixon Sep 24 '13 at 20:07
14

Images are indeed quite problematic when you need to scale them across different platforms with different pixel densities and different screen sizes.

Coupled with this, there are often problems with image manipulation libraries not being portable - especially when they use hardware acceleration on each platform.


Assuming that your rolodex app is somehow going to allow users to capture images, and then to upload them to some shared database/service for later viewing, here's how I might approach the problem. It's not the only solution - there's no "one right way" here!

** Capturing and uploading images **

  1. For capturing the picture (either from a folder or from a camera) I would have to use a native hook for each platform - so NOT PCL for this part

  2. If I then needed to do some on-device processing of the image (e.g. resizing or adding a thumbnail) then again this would probably be done differently on each platform - so NOT PCL.

  3. Once you have the photo (e.g. as a JPEG encoded in a MemoryStream) and want to upload it to a server, then I would probably do this using the asynchronous handlers HttpWebRequest (these are the old methods, nothing to do with async/await) in common PCL code

  4. What's running on the server.... well, that's almost certainly not PCL code - and I might use methods there to resize the image into device ready sizes - e.g. different size thumbnails

** Showing images **

  1. When I need to display someone's image from the database, then I would probably get the server to return a list of available thumbnail URLs - this is server code so would be not PCL

  2. The app code that actually asks for the list of contacts or the contact detail - that would probably be PCL code - and would probably use HttpWebRequest again.

  3. The View code that takes the contact and renders it on the screen? That would probably be XAML - and I would just use the native Image control on each platform to consume and render an appropriate thumbnail for the situation.

** If you were storing the images locally **

  1. If you were storing the images locally instead of using a central server, then you'll probably need to use non-PCL code for this too. Each platform has the same basic type of methods: LoadFile, SaveFile, etc - and each will provide mechanisms to create, enumerate, read and write folders and files, but each platform does this via a different API to the file system (e.g. System.IO in WPF, IsolatedStorage in WP7 Silverlight, etc).

  2. Once you've got your files into a common(ish) structure, then the PCL control code will be able to treat each file as just a pair of strings - the folder it's in and the file name...

  3. ... and in your UI layer (e.g. XAML) you will probably be able to reference those image files directly - probably can be done using a specific ValueConverter to generate the correct filename or filestream on each platform - e.g. in wp7 you could use the converter from Windows Phone 7 Silverlight binding image from the IsolatedStorage


So, my summary is:

  • I'd use PCL code wherever I could
  • but in reality that PCL code is only going to be in the control and networking - not in the image collection, image processing or image rendering

Some other things that might help:

  • currently I find it's quite hard to mix .Net4.5 async/await code with "normal" code - so even sharing the networking code might be quite hard to do - you might actually save time if you write separate code for everything :(
  • to help share the code, I would definitely try to use some form of IoC or DI to try to inject specific implementation of platform specific code where it's needed (this is what I do a lot in MvvmCross - and it's what @dsplaisted talks about in service location in http://blogs.msdn.com/b/dsplaisted/archive/2012/08/27/how-to-make-portable-class-libraries-work-for-you.aspx
Community
  • 1
  • 1
Stuart
  • 66,722
  • 7
  • 114
  • 165
  • I wanted to mark this as the right answer, it was hard to choose since this was a great answer. I just want to say thank you for the help, I can tell you spent a fair bit of time on the answer it it was very helpful architectural wise bringing up a few things I should keep in mind. I did upvote it :) Thank you! – Iris Classon Sep 20 '12 at 19:43
2

I tend to agree that image handling should probably be platform specific. Just as a general rule anything view related should not be included in a PCL. The above answers have some great technical detail that I will not bother repeating, but wanted to post a link to a slide deck I put together to discuss principals of using PCL. http://prezi.com/ipyzhxvxjvp-/sharing-is-caring-code-reuse-in-xaml-applications/

Hopefully this gives some general guidance on the overall design strategies and reinforces the information in the previous answers.

ITLackey
  • 21
  • 1
2

Most people have already said and the answer is also "Not possible". UI elements in PCL goes against the philosophy of PCL, which should be usable as is across platforms that you are targetting.

Here's what MSDN has to say on PCL - "the Portable Class Library project doesn't contain any UI components because of behavioral differences between the UIs of different devices" (from http://msdn.microsoft.com/en-us/library/gg597391.aspx)

hope this helps

Atul Gupta
  • 21
  • 1
0

My first instinct is to say no you shouldn't have images in your PCL given what you have described.

But if I were going to do it images can be stored via resources in the different libraries and you can store a different image for each platform in there and then do an Environment inspection to determine what image to pull out. But this approach really only should be attempted if you need scale.

By scale I mean you are open sourcing this library where 1000's of programs are going to use it. If this is for your own internal use for 3 or 4 apps, I would say don't bother save your self the headache. Just pass the image into the shared library or set it via a config and be done with it.

Or just don't manage images via the PCL. Because image manipulation changes and is different on every platform, especially the ones you mentioned. And you are bound to run into some weird bug where it just doesn't work the same as it does on the other platforms.

Nick Berardi
  • 54,393
  • 15
  • 113
  • 135
  • Do you know how to reference images in a PCL from a WinRT app? I'm just curious if that even can be done. Finding lots of people asking that, but no answers in RE to images – Iris Classon Sep 20 '12 at 12:48
0

I would probably put the actual image files in each app. I don't think you get much value out of trying to embed them in the PCL.

In your portable ViewModels, you can reference the image via a Uri. Unfortunately that Uri needs to be different on each platform. For Windows Store apps it needs to be something like ms-appx:////Assets/image.png where for Windows Phone I think it will just be /Assets/image.png.

So you'll need some code to figure out the correct Uri format. That could be a platform-specific service with a portable interface that the ViewModels would call. You might also be able to create a binding converter that would do whatever conversion you needed.

There are certainly other ways of doing this-- this is just what I'd picked based on the information in your question.

Daniel Plaisted
  • 16,674
  • 4
  • 44
  • 56
  • That's what i am currently doing, using an abstraction through 'adapters' that each platform send in to the ViewModel when DataContext is set. I guess a static class with a static BaseUri property could also work for this scenario, but I'm using adapters for working with files and storage etc. so I might keep that model just to stay consistent – Iris Classon Sep 20 '12 at 15:44
0

you can put the image on the view model as type "object" and use an interface (which is injected) to create the bitmap.

public interface IBitMapCreator
{
    object Create(string path);
}

public class MyViewModel
{
    private bool _triedToSetThumb;
    private readonly IBitMapCreator _bc;

    public MyViewModel(IBitMapCreator bc, string path)
    {
        _bc = bc;
        SetThumb(path);
    }

    public object Thumb { get; private set; }

    private void SetThumb(string path)
    {
        try
        {
            if (!_triedToSetThumb)
            {
                string ext = Path.GetExtension(path).ToUpper();

                if (
                    ext == ".JPG" ||
                    ext == ".JPEG" ||
                    ext == ".PNG" ||
                    ext == ".GIF" ||
                    ext == ".BMP" ||
                    false
                    )
                {
                    Thumb = _bc.Create(path);
                    OnPropertyChanged(new PropertyChangedEventArgs("Thumb"));
                }
            }
        }
        finally
        {
            _triedToSetThumb = true;
        }
    }

}

in the WPF version, you could do this

class BitMapCreator : IBitMapCreator
{
    public object Create(string path)
    {
        return BitmapFrame.Create(new Uri(path));
    }
} 

By using this method you can data bind right to the image and do not need a converter.

kns98
  • 330
  • 2
  • 7