There is a webpage which when loaded uses a random placement of forms / controls / google ads. However, the set is closed--from my tests there are at least three possible variations, with two very common and the third very rare.
I would like to be able to classify this webpage according to each variation. I tried analyzing the html source of each variation, but the html of all the variations is exactly the same, according to both Python string equals and the Python difflib. There doesn't seem to be any information specifying where to put the google ads or the controls.
For an example, consider a picture with two boxes, a red one (call it box A) and a blue one (call it box B). The boxes themselves never change position, but what takes their position does.
Now consider two possible variations, one of which is chosen everytime the webpage is loaded / opened.
Variation 1: Suppose 50% of the time, the google ad is positioned at box A (the red one) and the website control is thus placed at box B (the blue one).
Variation 2: Suppose also 50% of the time, the google ad is positioned at box B (the blue one) and the website control is thus placed at box A (the red one).
So if I load the webpage, how can I classify it based on its variation?
If the HTML is definitely the same every time, the variations are probably being done on the client side using javascript.
The answer depends on what you mean by "classify." If you just want to know, on any given load of the page, where the widgets are, you will probably have to use something like Selenium that actually opens the page in a browser and runs javascript, rather than just fetching the HTML source. Then you will need to use Selenium to eval some javascript that detects the widget locations. There is a selenium module for python that is fairly straightforward to use. Consider hooking it up to PhantomJS so you don't have to have a browser window up.
Related
I am automating a third party application with pywinauto. I am able to see some specific elements and their texts when using e.g. inpect.exe, but when looking at pywinauto's print_control_identifiers these elements are not included at all (marked with red in the image between the two green ones).
print_control_identifiers VS Inspect.exe (red ones are missing)
Is it somehow possible to see these elements with pywinauto?
In this particular case I need to check for example if an element containing the text "Varoitus!" exists, but for some reason pywinauto can't see it at all.
I'm using:
ctypes.windll.user32.SystemParametersInfoA(SPI_SETDESKWALLPAPER,
0, "picturefile", 0)
To change the wallpaper.
But I'm wondering if there's any simple way to put different wallpapers on each screen.
This feature isn't standard in windows though, but there are external applications like ultramon that do this. Anyone know how that works?
The way I thought it might work if I join the two images together into one and then make that the wallpaper, but then I still need a way to span one image accross both screens.
Also, how could I grab some info about the monitor setup, the resolution of each screen and their placement? Like what you see in the gui display settings in windows, but in numbers.
After joining the images together into a big image, you have to set the wallpaper mode to tiled to make it so the image spans the desktop (otherwise it will restart on each monitor).
Couple of ways to do this:
a) Using IActiveDesktop (which does not require Active Desktop to be used, don't worry). This is nicest as on Win7 the new wallpaper will fade in.
You create an IActiveDesktop / CLSID_ActiveDesktop COM object and then call SetWallpaper, SetWallpaperOptions and finally ApplyChanges. (As I'm not a Python dev, I'm not sure exactly how you access the COM object, sorry.)
OR:
b) Via the registry. This isn't as nice, but works well enough.
Under HKEY_CURRENT_USER\Control Panel\Desktop set:
TileWallpaper to (REG_SZ) 1 (i.e. the string "1" not the number 1)
WallpaperStyle to (REG_SZ) 0 (i.e. the string "0" not the number 0)
Then call SystemParameterInfo(SPI_SETDESKTOPWALLPAPER...) as you do already.
.
By the way, the code I'm looking at, which uses IActiveDesktop and falls back on the registry if that fails, passes SPIF_UPDATEINIFILE | SPIF_SENDCHANGE as the last argument to SystemParameterInfo; you're currently passing 0 which could be wrong.
EnumDisplayMonitors is the Win32 API for getting details on the monitors, including their screen sizes and positions relative to each other.
That API returns its results via a callback function that you have to provide. (It calls it once for each monitor.) I am not a Python developer so I'm not sure how you can call such a function from Python.
A quick Google for "Python EnumWindows" (EnumWindows being a commonly-used API which returns results in the same way) finds people talking about that, and using a Lambda function for the callback, so it looks like it's possible but I'll leave it to someone who knows more about Python.
Note: Remember to cope with monitors that aren't right next to each other or aren't aligned with each other. Your compiled image may need to have blank areas to make things line up right on all the monitors. If you move one of the monitors around and do a PrtScn screenshot of the whole desktop you'll see what I mean in the result.
I'm developing a python plugin for Rhythmbox - this contains a GtkScrolledWindow with a child which is a GtkIconView. The IconView is fed from a Gtk.TreeModel.
It looks like this:
Currently - and somewhat inefficient, every icon is drawn for every row in the tree-model - each icon is a GdkPixbuf from a file. If you have thousands of rows, it can take quite a while for the whole iconview to be fully updated with every picture icon.
What I am trying to achieve is to only update the icons that are in the current drawing area - when the user scrolls and releases the scrollbar (or navigates via the keyboard), the icons in the new drawing area should be updated with relevant pictures.
N.B. - the tree-model would be fully populated at this point - only the icons would not have been loaded.
This is not really my area of expertise - I'm looking for pointers for the best approach I should use to achieve the above.
Specifically - which Gtk+3 drawing-area signal (or signals) can be exposed (Gtk.ScrolledWindow / Gtk.IconView ?) to write python code to calculate what icons should be updated?
You should profile you application to see what takes time.
Is that loading the images ? If it is, then loading a default image and adding it everywhere in your view would be quick enough, as you'd load only one image. You'd then load and update the images on-demand using idle_add, based on the images that should appear in the viewport.
If what takes times is adding the images to the model, then you'd need to do the adding on-demand, by checking what is visible on the viewport in your idle_add callback.
If both are slow, you'd need a mix of both solutions: loading and adding on-demand.
Think also about the proxy design pattern that can be useful to create a fake cover object that will load in the background, and contain the loading policy.
For the signals, your GtkIconView widget implements GtkScrollable, which explains how to implement scrolling. You'd set your vertical adjustment and check when it has changed by connecting to its value-changed signal. This would mean the user scrolled up or down, and you'd need to fire up a timer with timeout_add. If after a short timeout (between 0.5 and 1s I think, but needs testing), the adjustment hasn't changed, this means the user stopped scrolling, and you can update what is displayed. Otherwise, it would be updated during the scrolling, slowing everything down. You then just need to figure out how to find which items appear in the viewport, to update their cover.
I've never done this before though, but I know GTK a bit and just tried to figure out how it would be done, so read that with a bit of caution. Anyway, the answer to reactivity is "on-demand".
But on the other hand if you have to scroll completely down, you would have to wait for every page to be build. So i guess you should better consider the idle_add function of the event loop.
This way your screen doesn't lock up and your pc doesn't have to wait for the user to load the view page for page. Win-Win for you and your application. ;-)
I'm new to QT. I'm using PyQT for GUI development in my project. I want to achieve this layout in my application. This application searches images from an image database. Google image search layout is ideal for my purpose.
I'm following the book "Rapid GUI Programming with Python and Qt" and I'm familiar with layouts. I guess I need to use a grid layout with each result image in each box of grid. & use vertical layout for (image,Qlabel,Qlabel) inside each grid box.
These are some problems I'm facing.
Importantly, I'm unable to display image. What control/widget do I need? I cannot find anything similar to PictureBox of .NET
How do I seperate these image result by fixed gap like in the image? I'm using Horizontal & vertical spacers but that isn't working?
How to set QLabel a clickable (like hyperlink). I don't want to a open a URL. Just the text should be clickable. So, that when user clicks on the link. I can show more information (like next set of results when he clicks on next page number or a new window with image in fullsize when user clicks on 'view') Do we have some new kind of control for this?
This is another important issue. I'll display the page numbers of results (like shown in figure) & assuming they are clickable. How do I load a new page of results? I mean what is the equivalent of page in QT?
As you can guess. This definitely wont be the first page of GUI. The first page will be exactly like http://google.com (a big logo & text box with button below it). when user clicks the search button. This page will be displayed. Again the same question comes up. How change the pages?
Please give a list of controls I'm going to need for this. Tell me if I'm unaware of something.
1/2.
For displaying the images and labels use a QListWidget with view mode set to QListView::IconMode. However, if you need to customize the display beyond what the QListWidget/QListWidgetItem api can provide you will need to create your own QAbstractListModel and use a standard QListView with it.
Make sure and read Qt's primer on model/view.
As for spacing the images, checkout the spacing property on the list view.
Here is an example from KDE's Dolphin file manager:
3. Use a regular QLabel, but set the contents to be an href.
Example:
edit: Oops I see from your tags you are using PyQt, the following is C++, but should be similar to what you would do in python.
QLabel *linkLabel = new QLabel;
linkLabel->setTextFormat( Qt::RichText )
linkLabel->setText( " Click me! " );
connect( linkLabel, SIGNAL( linkActivated ( const QString & link ) ), .... )
4.
Well, since you are using a Model/View, why bother having page numbers at all? The user will just be able to scroll the view and more pictures will be shown. This is by far the easiest solution as you don't have to do anything once you've got your M/V setup!
However, if you really want to show page numbers it will require more work in your model. For example, have a track the "current page" in the model and only allow access to images on the "current page". Then in your slot connected to the linkActivated() signal tell the model to change pages. I won't go into much more detail as this seriously violates the whole idea behind model/view. The "right way" of doing this would be to subclass QListView and add pagination support, but like I said why not use scroll bars? There isn't any performance hits to doing so.
5. Use a QStackedWidget, addWidget() all your "pages" to it, then call setCurrentIndex/Widget() as needed to switch the pages.
Thoughts:
It seems you are very committed to cloning the look, feel, and behavior of Google Image search, which is fine, but Google Image Search is a web application that uses interaction paradigms that are very different than a normal desktop application (links, pages, etc). You are presumably developing a desktop application, and by trying to emulate the behavior of a web app you will find it difficult as the API just isn't designed to support those sorts of interactions. By all means, it is doable, but you'll have your work cut out for you.
If you are extremely intent on sticking to the web based interaction style, why not code your app in javascript and HTML and toss it in a QWebView?
Try using QListWidget with viewMode set to IconMode. It should do all for you. BUT if you need to customize your data display use QListView with your own/standard model and own delegate for painting
I'm not familiar with PowerBuilder but I have a task to create Automatic UI Test Application for PB. We've decided to do it in Python with pywinauto and iaccesible libraries. The problem is that some UI elements like newly added lists record can not be accesed from it (even inspect32 can't get it).
Any ideas how to reach this elements and make them testable?
I'm experimenting with code for a tool for automating PowerBuilder-based GUIs as well. From what I can see, your best bet would be to use the PowerBuilder Native Interface (PBNI), and call PowerScript code from within your NVO.
If you like, feel free to send me an email (see my profile for my email address), I'd be interested in exchanging ideas about how to do this.
I didn't use PowerBuilder for a while but I guess that the problem that you are trying to solve is similar to the one I am trying to address for people making projects with SCADA systems like Wonderware Intouch.
The problem with such an application is that there is no API to get or set the value of a control. So a pywinauto approach can't work.
I've made a small tool to simulate the user events and to get the results from a screencapture. I am usig PIL and pytesser ORM for the analysis of the screen captures. It is not the easiest way but it works OK.
The tool is open-source and free of charge and can be downloaded from my website (Sorry in french). You just need an account but it's free as well. Just ask.
If you can read french, here is one article about testing Intouch-based applications
Sorry for the self promotion, but I was facing a similar problem with no solution so I've written my own. Anyway, that's free and open-source...
I've seen in AutomatedQa support that they a recipe recommending using msaa and setting some properties on the controls. I do not know if it works.
If you are testing DataWindows (the class is pbdwxxx, e.g. pbdw110) you will have to use a combination of clicking at specific coordinates and sending Tab keys to get to the control you want. Of course you can also send up and down arrow keys to move among rows. The easiest thing to do is to start with a normal control like an SLE and tab into the DataWindow. The problem is that the DataWindow is essentially just an image. There is no control for a given field until you move the focus there by clicking or tabbing. I've also found that the DataWindow's iAccessible interface is a bit strange. If you ask the DataWindow for the object with focus, you don't get the right answer. If you enumerate through all of the children you can find the one that has focus. If you can modify the source I also advise that you set AccessibleName for your DataWindow controls, otherwise you probably won't be able to identify the controls except by position (by DataWindow controls I mean the ones inside the DataWindow, not the DataWindow itself). If it's an MDI application, you may also find it useful to locate the MicroHelp window (class fnhelpxxx, e.g. fnhelp110, find from the main application window) to help determine your current context.
Edited to add:
Sikuli looks very promising for testing PowerBuilder. It works by recognizing objects on the screen from a saved fragment of screenshot. That is, you take a screenshot of the part of the screen you want it to find.