Re: what's new in 8.0.1?
- To: mathgroup at smc.vnet.net
- Subject: [mg117329] Re: what's new in 8.0.1?
- From: David Bailey <dave at removedbailey.co.uk>
- Date: Tue, 15 Mar 2011 06:05:24 -0500 (EST)
- References: <ilksj9$6ff$1@smc.vnet.net>
On 14/03/2011 11:02, Bill Rowe wrote:
> On 3/13/11 at 5:24 AM, mike.honeychurch at gmail.com (Armand Tamzarian)
> wrote:
>
>> On Mar 12, 9:08 pm, Bill Rowe<readn... at sbcglobal.net> wrote:
>>> On 3/11/11 at 4:35 AM, mike.honeychu... at gmail.com (Armand
>>> Tamzarian) wrote:
>
>>>> 1. re: bug. Evaluating user documentation before a release ought
>>>> to be part of the pre-release screening. Bugs that show up when
>>>> evaluating user documentation are, by definition, bugs that are
>>>> readily detectable. Allowing these through creates perceptions of
>>>> carelessness and poor quality assurance.
>
>>> How easily an issue in user documentation is found clearly depends
>>> on the size of the documentation. For Mathematica version 4 (when
>>> the documentation was still be distributed as hard copy), the
>>> documentation was a 1470 page hard bound volume with an additional
>>> 535 page soft cover volume. For Mathematica 8, if the documentation
>>> were printed, it would likely be 15-20 volumes of the size of the
>>> hard bound volume that came with version 4. It is unrealistic to
>>> expect that much documentation to be error free. Nor is it accurate
>>> to say errors in that much documentation are "readily detectable".
>
>> Couldn't disagree more. Evaluate notebook and then search for pink
>> boxes, search for error messages. Are you suggesting that couldn't
>> be done programmatically by the makers of the best software ever
>> invented.
>
> Apparently, we have different ideas as to what constitutes
> errors in the documentation. What you suggest above would only
> catch problems with the supplied examples. It would not find
> problems where the documentation indicated a different behavior
> that actually occurs. I agree for a suitably restricted set of
> errors, they should be readily detectable. But I definitely do
> not agree *all* errors in the documentation are readily detectable.
>
>
I have always understood that WRI has a large database of tests that it
executes EVERY NIGHT to ensure that their internal version is valid. I'd
have thought adding the documentation examples to that database should
be relatively easy to automate, starting from the input cells, and the
contents of the corresponding output cells in the help notebooks -
presumably created at the time the examples were created.
Obviously, the first time this was run, various spurious errors would
pop up, such as Module[{x},x] (where the unique suffix applied to x
would vary), but after those glitches were removed, the test could be
totally automatic.
Of course, sometimes a help example fails because of some circumstances
that are special to a particular installation, so even this would not
achieve perfection.
David Bailey
http://www.dbaileyconsultancy.co.uk