Quoting hydr0x:

Quote:
It is not. Just half it every time. Select the first 14k profiles, and export. If it works, the issue is with the other 14k. The 14k that fail, split into 2x 7k and try the export. Repeat until done. It takes 15 such splits to determine the problematic profile.

Funny, I just had the same idea. And it might be possible to partially automate the process. One could write a program that keeps track of the tested parts and flags the next part to be tested, and then starts the export. The only manual input would have to be "Did this export work? Yes/No". I'll discuss this with mediadogg.

If you can export your collection into two separate export files, then you can combine them into a single export file. Theoretically you can do it with a good text editor, but with such big files it's tricky.

As luck would have it, I have a program that merges two export files into one. Originally this was intended to merge a small export of flagged profiles into a master export file, so I wouldn't have to do a full export to get my export file up-to-date. However, it works fine for this purpose as well.

It's a command line program, but I have written a user interface for it.


In the example above, it would merge Export2.xml into Export1.xml, so that Export1 would be the complete export file.

If you would like to try it, you can get it here.

Well, the corrupt part 2 might possibly give you a clue as to where the problem lies.
I have a tool for analyzing broken export files; DpXmlAnalyzer.

It can give you something like this:


Hopefully the first error that it finds points to the problem profile. No guarantee, but worth a try.

If it works without an error message, then that means that the file contains all valid XML. So no help, I'm afraid. A closer inspection of the file might yield something. Maybe mediadogg has some idea …

By the way, I might have discovered the root cause of Magmadrag's issue. I have found that even when I explicitly create a cast entry with BY 0, for a cast member that already exists with a non zero BY, it doesn't work!!!!

I made a very small test database and started from totally empty. It seems that when i do it with the API, the profile gets the BY 0 entry, which makes  it look like things are good on the surface. But in fact, the BY 0 entry does not get captured in the Actors table, and Profiler assigns the credit from the matching name with a BY - the very issue Magmadrag is trying to work around.

The only way I could do it, was manually using the GUI. In edit mode, I could create a BY 0 entry and everything works as expected. You get the Actors table with both entries. So we are dealing with an Invelos bug - an API bug that will not allow you create a BY 0 entry for a cast name that already exists in the database with a BY.

I will continue my testing and I hope my above conclusion is wrong.

I wonder if DJ Doena has something to say about this - with all of those Cast / Crew tools, what have you found?

I have an idea. I am going to create an entry first with a crazy fake BY, see if that works. Then try to modify the profile and change it to BY 0. I wonder ...

Well friends I might have hit an untimely and unexpected demise of this concept. The API will not let me stuff a zero into BY. I can put any other crazy number, even negative! And that number (or its compliment if negative) winds up as a new entry in the Actors table and also in the profile. Soon as I try to do it with zero - nada. The BY always comes back with the original, non zero entry. And the BY 0 entry is not saved.

But if you do it by hand, works Ok, which Magmadrag knows well, because that is what he is doing, painstakingly, one by one.

The only thing I can think of right now is some kind of AutoIt augmented process that feeds off of a list of candidates that we know how to generate using the famous "Magmadrag Method." 

Any other ideas?

Well that kind of works, except the plot thickens. It seems you can't always believe what you see. Profiler seems to remember credits that have been deleted in some cases. Then just when you thought you knew what was going on, the things get corrected. Like a delay in registering changes to the database.

I am so dizzy right now, I need to take a break. I am convinced my code is doing what you want want it to do, on the surface, but the results you see on the Profiler screen are sometimes confusing and not always consistent with what you just told it to do. Sometimes things get corrected when you close and reopen the profile editor, and maybe other cases I will discover.

The API has some functions that address database sync issues, and after a billion hours, I got it working in BulkEdit. But even sometimes when I directly copy / paste code snippets from there, that work, they don't work in the new place. Is it me, my age, my impatience, my fading brain ...

My programs also need 4.7.2. I don't know what you are doing, but the higher versions of .NET should be backwards compatible. Also, you should not have to uninstall prior versions.