I might mention that I don't think that a tag can be added if it has been previously deleted (by a moderator, say).

Sometimes people add tags which are so vague or ambiguous that they're meaningless, for example, "process", or "problem", or "math".

Or they add a duplicate tag which is very, very close to some other, ie. a close variant or misspelling. For example, "images" versus "image".

Or they add a tag which is overly obscure (in the sense of being rare jargon), ie. too obscure to be generally useful and generally recognizable, eg. "enthalpy-of-sublimation".  It's a problem because the system sometimes suggests "close" tags to posters.

nb. I've deleted many poor tags, over the decades. (Actually, there used to be a small number of members -- no longer so active -- who added poor tags like it was going out of style.)  But so it goes...

The "maple" tags for stackoverflow and match.stackexchange have also seen almost complete drop off in new Questions, since about the middle of 2025. I have answered many/most of the answered questions with that "maple" tag, for the past 15 yrs or so, on those two sites, and the trend was hard to miss. And the calendar dates of Questions with that tag shows the remarkable dropoff that began near the end of 2024 or so.

As far as I can tell the level of new general non-Maple-specific Question activity on stackoverflow has dropped like a stone, since the advent of stronger generative AI.

So, the Mapleprimes questions have not gone to stackoverflow or math.stackexchange, it seems.

And afaik people are still purchasing and using Maple, MapleSim, Maple Flow, Maple Learn, Maple Calculator, etc.

I believe that LLMs can and have been trained wrt Maple in part using what they've scraped off of this site, or the Online Help, or other forums, or from accessible books. (Not all, but a great deal of the inlined code here is in 1D plaintext.) The data must be out there, because some of the models are quite adept at Maple.

I think that many questioners have simply moved on from here; now they mostly ask AI, either directly or through Maple 2026's MCP, etc.

@Rouben Rostamian  I am not sure how to force one of those forms to be manipulated into the other.

But here is a forced equivalence check: Maple 2024 like yours(?), but also for Maple 2026.1,

Download mw_RR1_ac.mw

There is also this form...

simplify(expand(ans_tmp));

     1/4*(3 - 2*2^(1/2))^n
   - 1/2
   + 1/4*(3 - 2*2^(1/2))^(-n)

Download fun_acc.mw

Also,

simplify(numer(expand(rans - ans)));

          0

simplify(denom(expand(rans - ans)));

    4*(3-2*2^(1/2))^n

ps. I can submit a bug report against that as a weakness in simplify.

@Mac Dude I previously gave a terser template for dealing with that programmatically.

Turning that into a simple operator is easy enough:

Download Maple_freeze_ac.mw

As you've doubtless realized, this is not going to always be The Best Way, for all examples. But it's not a bad addition to the tool bag.

@Mac Dude The issue of whether the sqrts' values are real or imaginary (under assumptions, or whatever) is not a part of what's going on in this example.

There's no combining of sqrts, either with other radicals or wrt to cancellation other factorizable expressions. That's why the symbolic option of either simplify or combine is not the key bit of missing magic here. The sign of what's inside the radical, or branch cut stuff, is just not the difficulty here.

That's why the target rearrangement is tricky to obtain even after one replaces all those multiplicands (including radicals) in the denominators by mere names. See my followups above; that's why I showed them. The issues exist even for expressions that essentially can be represented as one polynomial (of rational coefficients) over another.

And assumptions are directly unrelated here, in consequence.

Rather, your example's issues mostly boil down to just a few things, such as,
1) simplify(...,size) tries to be fast, and so doesn't attempt all possible rearrangements such as those doable via collect (or via other arithmetic operator rearrangements, etc).
2) collect won't collect wrt all things; it can baulk on radicals, reciprocals of radicals, other reciprocals, and so on.

I've been on this ride several times before, over decades. Many years ago I wrote myself a super-compressor utility, that speculatively tried all manner of arithmetic and term-wise rearrangements and factoring and collecting and what have you. And it could make a deep nesting of such command calls. Quite often it beat simplify(...,size), since it didn't have to be fast. (I also wrote it to spit back the sequence of command steps that could reproduce the result.) It's important to remember that getting a completely optimal result can be relatively quite expensive.

Note too that your target form is not optimal in terms of expression "size". It's just quite good. You may get various suggested Answers (some ok, some not ok). But if you're shooting for a specific target form (that might not be optimally "small") then you may need to characterize it's form and nature -- else how could one reliably program for it...

ps. Please understand that the code nm posted is pretty severely flawed. It got your particular example ok, but it also produces completely wrong and invalid results for a large swath of other examples -- some of which are not too far from the form of your example.

I remembered that I'd deliberately used C rather than 1/C for the substitutions above.

But it's also interesting to substitute for 1/C, similarly to how the radicals (power negative 1/2) are treated above.

And we can also notice some variants due to different ordering and simplifier of the collect call.

nb. I have previously noted, elsewhere, that simplify(..,size) doesn't try all (or maybe even any) possible collect choices, speculatively. It's a generally expensive thing to do, blanketing the choices. (I have a crusher that tries them, but in general it's not fast.)

Download Maple_example_ac3.mw

@nm I see at least five programming problems with that my_simplify procedure.

Alas I don't have time to describe them all. I'll just show that for several kinds of expressions it produces results that are not mathematically the same (or even close) to the original.

Download maple_sol_acc.mw

I'll try to explain my line of attack.

Download Maple_example_ac2.mw

An initialization file can go inside special locations under the folder in which Maple itself is installed.

It can also go inside some special location under a user's personal config files on the OS (the location of which varies by OS).

It is explained in a Help page, the online version of which is here. (That Maplesoft Online Help site might be having certificate issues at present, requiring http rather than safer https access at the momment. The page is also available in Maple own Help browser, under the Help Topic worksheet,reference,initialization  )

ps. some of the replies above show some misconceptions in this (and other) regards.

@Ronan I only suggested that you might not have to add to libname.

And that indeed was your question: "I need to add a library path name to get an old style package .lib to run. Currently a have to manually add the path. How do I solve this?"

I didn't suggest that you do something new/different wrt savelibname. You might need to revise your assignment to it, accordingly (or possibly not assign to it at all, perhaps).

It's quite possible that something inside that eClifford archve package does something with savelibname. If that were the case then I likely wouldn't be able to change what it does.

So you might have to ensure that it were (now) assigned to match the new location that ended with /lib .

It's even possible that something in the package does something crazy and near-unworkable. I wouldn't know what it does, how it was built, what it's internals do (maybe something old-fashioned and no longer necessary such as call readlib), etc.

ps. I don't know why you had savelibname assigned, at all, and you haven't explained why.  Was it part of some magic just to get this eClifford package to run, earlier? Or was it something you do for your own package development? If the latter then I strongly advise not assigning and using savelibname (it's a horrid concept). Instead, I'd just use LibraryTool commands and sane, explicit reference to target archive files.

@C_R Your approach does not account for the fact that the result only needs to agree up to a piecewise constant.

For real x, see the last call in my earlier Reply, where it differs by Pi*I/2.

For method=risch, you can see something where it differs by not just a constant but by a piecewise constant (even for real x).

int(D(tanh)(x),x,method=derivativedivides, method=risch);

   -2/(exp(2*x)+1)

plot(% - tanh(x), x=-7..7, adaptive=false, numpoints=400, style=point);

That's why vv differentiated, before his check. (That's a usual check, since the additive constant disappears.)

@Kitonum Yes, style=surface is a usual way to beautify the implicitplot3d result. Vote up.

Since it's a mere plane, is it not the case that adding the option grid=[2,2,2] for the implicitplot3d call makes it faster to compute and with a much smaller size structure.

@dharr For the range under discussion, could one easily and programmatically get to, say,
    (25*x*y + 5)^(1/5) = (x - y)^(1/4)
or,
    25*x*y + 5 = (x - y)^(5/4)

I mean, we could divide sol2 through by 7, etc.

@Rouben Rostamian  

Download test_AF2_nm_ac.mw