That's a pretty general statement, so I'll give you a specific example. One of the great features of LabVIEW is the Context Help system. The information you see, when you press and move the mouse over a block diagram object, is determined by three fields in the VI Properties dialog box: VI description (the actual text displayed below the connector pane image), Help path (the path to the .CHM file that contains the help topic), and Help tag (the HTML file name contained inside the .CHM file). When the last two fields are valid, the Context Help window displays a "Detailed Help" hyperlink below the VI description.
We have internally-developed tools to "populate" (import) the Context Help -- by which I mean, transfer the proper information from our source material into the VIs themselves -- before release, so all the information appears properly in the software. But (like all software) these tools aren't perfect, so we have more processes in place to verify a successful Context Help population before release.
Now, LabVIEW has a HUGE library of VIs, functions, structures, constants, and so on -- and almost every object has controls and indicators with even MORE context help. Not to mention dialog boxes, which contain even more information. I don't know exactly how many, but the core product has tons, and our add-ons add on (hah!) even more. This massive number makes it really challenging to manually verify the Context Help population before each release. And yet that's what we do -- spend a lot of time hovering over icons and clicking Detailed Help links, and filing bug reports when necessary.
Is this starting to sound familiar? I bet many of you reading this blog have procedures for manually testing your products. Wouldn't it be great if there was a way to automate these tests? Funnily enough, I work for a company that develops products to do just that ...
So what I'm in the middle of doing now is developing an automated testing system, using LabVIEW, to verify the context help integrity of any given number of VIs. You give it a top-level palette (.mnu) file, and the VI I'm writing will recurse (yes, recurse!) through all the subpalettes, build a list of items on those palettes, and verify the Context Help for each item.
No more hovering over palette items ad infinitum. No more manual testing. You just specify a .mnu file and away the program goes. And in fact, since the path to the .mnu file is not likely to change from day to day, you can run this VI programmatically by passing in the path as part of another process. And there you have it, a fully automated test. All you do is check the report when you come in each morning. "What errors do we have today?" And you can use this information to improve the tools that actually import the context help, because you discover all the little corner cases that you missed when developing those tools.
This is just one example of internal tools development. We've developed a whole suite of them, all using LabVIEW, to verify the integrity of our HTML help files, and use them every day. Just as the software itself might build automatically every night, so do our .CHM files. And just as software autotests might run every night while you sleep, so do our .CHM tests.
I say we really put the "technical" in technical writing.
Sounds like fun, could it be that such tools are released?
ReplyDeleteTon
It appears that 'eating your own dog food' is becoming popular at NI.
ReplyDeletehttp://www.shanghaitechwriter.com/2008/03/31/eating-your-own-dog-food/
LabVIEW using LabVIEW is always a good thing. Please pass the dog food!
ReplyDeletePhil,
ReplyDeleteAbsolutely! We've been doing it for a long time here, especially with TWs who have CS backgrounds. So we've built up a lot of LV programming experience on the technical writing team.
An engineering mind,
ReplyDeleteFor me, a really fun part is using our own LV documentation to help me create VIs -- then realizing that the person who wrote that help file is sitting like ten feet away :-)
tonplomp,
ReplyDeleteAnything's possible :-) I haven't given that option a lot of thought, but if I can get everything working properly, who knows?