Hello Tomaž
I have been exploring today, to try to narrow down where the problem appears. It’s difficult to give you some code to reproduce the problem, because you would need to have my Outlook Data File, or an extract of it, plus MS Outlook to run it on. If you want that, I will try to work out how to produce a minimum extract to show the problem. Meanwhile, I can describe what the problem seems to be. All my work has been done in a playground. I can set up a connection to my Outlook Data File and navigate through it to the folder of interest, which contains editions of a newsletter in German. I then inspect one of the items in the folder, which is a COMDispatchInstance and an instance of the object MailItem. My interest is the property HTMLBody of the MailItem, which is as you expect the complete code to display the newsletter. If I inspect the HTMLBody property for most MailItems, it displays with no problem. In some cases, however, as soon as I click ‘inspect it’, there is a message saying the VM has crashed, and the system closes with a crash dump. I don’t think there is a problem with the fact that I am displaying the property value in an inspector. I tried instead to assign it to a variable in the playground. As soon as I clicked ‘do it’ the whole Pharo app closed down, with no message or crash dump. I have tried to find some common factor in the items which will cause the crash, and the one thing I can find is the size of HTMLBody, or equivalently the size of the MailItem. The HTMLBody displays as a WideString in the inspector, and it works fine for an object of about 230K characters. With another message about 10% larger, the VM crash occurred. I have no knowledge of COM interfacing, except what I have picked up in the last two days, but I have been trying to find a general description of a problem which would give the results I have seen. If a COM object has a property which is a BSTRString, and if the string is a WideString with more than about 250K characters, is there some internal buffer overflow? – the crash dump has several mentions of ‘Stack Overflow’, though I can’t really follow it. That is about all I can explain – sorry it’s a bit long-winded. If you want more information, or an extract of my Outlook data file, just let me know. Thanks for your help. Peter Kenny From: Pharo-users <pharo-users-boun...@lists.pharo.org> On Behalf Of Tomaž Turk Sent: 07 April 2020 16:30 To: Any question about pharo is welcome <pharo-users@lists.pharo.org> Subject: Re: [Pharo-users] Automation of MS Office from Pharo Dear Peter, I'm glad that you find the package useful! Can you please explain a little bit more about your code, because it's not obvious from your description at which stage the problem occurs. Could you trace it with the debugger, and/or possibly post here the piece of code that fails? Thanks and best wishes, Tomaz ------ Original Message ------ From: "PBKResearch" <pe...@pbkresearch.co.uk <mailto:pe...@pbkresearch.co.uk> > To: "Any question about pharo is welcome" <pharo-users@lists.pharo.org <mailto:pharo-users@lists.pharo.org> > Sent: 6. 04. 2020 19:15:16 Subject: Re: [Pharo-users] Automation of MS Office from Pharo Hello Pablo (and Tomaž) You asked for feedback; after a day of experiments, I have good news and bad news. The good news is that I have been able to navigate through my Outlook Data File, find my way to an e-mail of interest and examine the fields of interest. This means I am 90% of the way to my objective. All I want is to get the HTML code from the message body and pass it to XMLHTMLParser for further analysis. I found it not too difficult to work this out; the fact that a COMDispatchInstance can display all its internals in a playground page meant that I could work it out interactively as a playground exercise. I am still discovering things about the interface; I have only just found out that I can select an item from a list by quoting its name as an argument, rather than its number in the list. I am sure my clumsy code can be tidied up still. I can get to the 100% point in the job by exporting the contents of the e-mail to disk in HTML format, using the SaveAs procedure of the MailItem object. I can then read it back from disk into the parser. But this seems a bit clunky. The content of an HTML mail item is available in the HTMLBody property of the MailItem, so I should be able to pass it programmatically to the parser without going near the disk. This is where I run into trouble. In most of the cases I tried, this worked fine; I could display the HTML body text as a WideString in the playground, assign it to a variable and do whatever I wanted. However, for just one e-mail I tried, as soon as I selected ‘Do it and go’ in the playground, a message came up that the VM had crashed; the application closed, leaving only a crash dump. I cannot find anything unusual about the message that failed in this way, except that when I save it, it produces a larger file than any of the others (just over 1MB, against up to 300KB for the ones that worked). Could there be a limit on the size of some internal buffer? I thought it worth while mentioning this now, because crashing the VM is generally undesirable. I can solve my problems safely using the SaveAs route, so it’s not a problem for me. Hope this helps; if you want more details, let me know. Overall I am very happy with this library. Peter Kenny From: Pharo-users <pharo-users-boun...@lists.pharo.org <mailto:pharo-users-boun...@lists.pharo.org> > On Behalf Of teso...@gmail.com <mailto:teso...@gmail.com> Sent: 06 April 2020 11:24 To: Any question about pharo is welcome <pharo-users@lists.pharo.org <mailto:pharo-users@lists.pharo.org> > Subject: Re: [Pharo-users] Automation of MS Office from Pharo Hi Peter, First, thanks to try to use Pharo-COM, that is great, I love to have users for it and find it is useful. Secondly, the Pharo 7 problem is an error I have introduced. It is clear that some changes I have done to support the new version of UFFI (the framework in Pharo to handle FFI calls) have broken the Pharo 7 version, so I will fix it to maintain compatibility. It is great that you were able to make it work! Cheers, On Mon, Apr 6, 2020 at 12:18 PM PBKResearch <pe...@pbkresearch.co.uk <mailto:pe...@pbkresearch.co.uk> > wrote: Hello Tomaž Many thanks for your patient explanation. I should have been able to work out for myself how the Word test works, and indeed I realized some of it when I came to shut down the Word instance and its two documents. But it was late at night, and I should have packed up before then. My last test last night showed that basically I have cracked it for my job of automating Outlook. I have been able to connect to my running instance of Outlook, open my application and interrogate the names of my top-level folders. From now on it should be just a matter of understanding the MS documentation of the Outlook model. However, all this is with Pharo-Com installed in a new Pharo 8 image. I have no idea what went wrong with my first effort on Pharo 7. But I shan’t worry about that – I shall gradually move all my bits and pieces to P8. I shall try to work it out myself from here, but I shall come back if I get stuck. Thanks again Peter Kenny From: Pharo-users <pharo-users-boun...@lists.pharo.org <mailto:pharo-users-boun...@lists.pharo.org> > On Behalf Of Tomaž Turk Sent: 06 April 2020 08:04 To: Any question about pharo is welcome <pharo-users@lists.pharo.org <mailto:pharo-users@lists.pharo.org> > Subject: Re: [Pharo-users] Automation of MS Office from Pharo Hello Peter, If you look at the code in the Word test you will notice that the test firstly - creates a new Word instance, - makes it visible to the end user, - then adds an empty document to the documents collection with the text "Hello from Pharo!" - then it tests whether it can receive the same text back from Word. After that, the test - adds a new empty document to the documents collection with the text "Hello from Pharo! Some additional text. ", this time as an array of two texts - it activates this second document (this imitates the end user's window activation on the desktop) - then it tests whether it can receive the same text back from Word. If you look at the Task Manager, you'll notice that you have one Word process with two open documents: Namely, for each document Word creates a new, separate window - the documents are not displayed in one "Word application window", but separately - that's a normal behavior for some versions of MS Office, and it happens also if you open several documents directly in Word. So, there is just one Word instance. 'finalize' clears the references to the Word instance, it doesn't close the program by itself. If you want to do that, you can send Quit message to Word before you destroy the reference (https://docs.microsoft.com/en-us/office/vba/api/word.application.quit(method)). Similar behaviour is with Outlook, here's one example: https://www.excelcommand.com/excel-help/excel-how-to.php?i=124116 The calling among COM objects is asynchronous, and it's usually wise to wrap it in error handling structures. Please tell us how it goes. Best wishes, Tomaz ------ Original Message ------ From: "PBKResearch" <pe...@pbkresearch.co.uk <mailto:pe...@pbkresearch.co.uk> > To: "Any question about pharo is welcome" <pharo-users@lists.pharo.org <mailto:pharo-users@lists.pharo.org> > Sent: 5.4.2020 23:18:02 Subject: Re: [Pharo-users] Automation of MS Office from Pharo Pablo - a final update before I close for the night. The Word test on the pharo 8 version comes up green. The strange error message is nowhere to be seen in any Pharo 8 runs. The result is not what I expected; I finish up with two Word documents open, one with the first message, the other with the two messages. I thought the 'finalize' command would close it down. Anyway, it looks as if I need to switch to P8 to use Pharo-Com. I shall continue testing tomorrow on P8. Sorry for the late-night hassle. Peter -- Pablo Tesone. teso...@gmail.com <mailto:teso...@gmail.com>
