I like the idea to have a free "Mathcad", but since I have compared the results with Mathcad, unfortunately I and can not trust it anymore . Attached there is a file showing the wrong result. I have exported the same file to mathcad and got the correct result there.

---

ftbktest.sm (13 КиБ) скачан 77 раз(а).

Wrote

I like the idea to have a free "Mathcad", but since I have compared the results with Mathcad, unfortunately I and can not trust it anymore . Attached there is a file showing the wrong result. I have exported the same file to mathcad and got the correct result there.

The unreliable symbolic engine is a major problem in SMath. The bug-tracking system is full of examples. Your example is particularly critical as there is no fair chance to recognize the error unless you double check with some other software. I added a link to your example in the corresponding bug report.

SMath is a mine field and you cannot operate safely without sticking to some survival rules. One of them is to set the optimization of all definitions/assignments to numeric or none (right mouse button menu).

There have been multiple requests in the past to set the default optimization for assignments to numeric or none (ss-65) and to make the setting visible for individual formulae, perhaps using distinct assignment and evaluation operators ss-19.

I tried to support the development and application by writing a handbook and submitting bug reports and feature requests. That, however, obviously did not stop you from stepping into the optimization trap and the forum is full of similar cases. SMath would need a splash screen saying: "Always make sure that your definitions are set to "optimization none" using right mouse button menu". Better, of course, would be to have this setting by default.

I have to admit that after having used SMath for teaching for three years, I now surrender and purchase a Mathcad classroom license for our CAD/CAE lab.

Unfortunately, I have to agree with Martin completely. I also think that Andrey did not expect such a great interest from the community and even greater expectations from the SMath users when he started this project. There is a bizarre thing regarding to SMath, as Martin mentioned. A Math software which is sometimes unreliable in calculating rather simple things . In my opinion, If there would have been a chance to fix this, Andrey would have done this after all these years. If SMath fails to perform reliably simple algebraic calculations with few numerical variables, sooner or later - nobody will us it.

Regrettably, if Martin surrendered - there is a red light for SMath . Something has to be done or Smath will die, and I think this is a fact.

Regards,
Radovan

Thank you Martin, very helpful, and I completely agree with Radovan!

Wrote

In my opinion, If there would have been a chance to fix this, Andrey would have done this after all these years.

I disagree. Yes, it might be hard to solve the problems with the symbolic engine. I am fine with Andrey waiting for a better world to solve the problem.

But it cannot be that hard to set the default optimization to none (if I can change the setting, the core program should be able to do so as well).

Also, the distinct operators for symbolic or numeric optimization should not provide prohibitive challenges. As you can see in the live version, there is a right arrow for symbolic evaluation and a = for numeric. This used to be this way in older versions of SMath and has been removed deliberately (Mathcad patent issues?). But again: The problem is in the optimization settings of assignments, not of the evaluation/display.

I guess that the symbolic engine is something Andrey has put into quite some effort and just cannot stand the idea of switching it off by default.

This reminds me of what is known as trap of emotional commitment in mechanical design. Designers tend to use 3D CAD models right from the start of a project. They focus on details, which is immediately rewarded by nice looking models. Questioning the basic concept or even the customer's requirements does not immediately pay but can avoid failure in the long run. The further you go in detailed design, the higher the costs of and the resistance to conceptual changes.

But perhaps it is not Andrey who stepped into the trap but me and Radovan and others. Once we went through the pain of finding out how to operate SMath, we tried to reward ourselves by sharing the word to others, collecting thanks in the forum. Do you remember, how long it took for you to find out how to write ranges a = 1..10? For me, that was several days. In order to not write off this as waste of time, I wrote the handbook. And so on. But doesn't the handbook contribute to the illusion (it is an illusion, given the recent bug report by Dijay) that SMath can be used for serious work? Isn't it a danger to the public if people start to use SMath for designing real things?

A program without an active user community is dead. In order to keep the community active, there must be some safe and reliable use cases. That is why we ask for default optimization settings being None or numeric. SMath can still be very useful without symbolic optimization of assignments.

Wrote

But it cannot be that hard to set the default optimization to none (if I can change the setting, the core program should be able to do so as well)...Isn't it a danger to the public if people start to use SMath for designing real things?

I totally agree to this also. To set an unrelieable calculation method as default is either stupid or for test reasons. I believe in the second and this might be the reason, why we only have a draft version 0.9x up to now - that's serious and not dangerous!

The risk is to damage the Name of this Project! The author simply should find a way to warn the People when using the symbolic engine!

If I'm right and the author is interested in testing results I propose another way: Set the numeric Modus as Default, install a test button for symbolic modus and automatic comparison of the results and aks the community to help testing!

Wrote

Wrote

In my opinion, If there would have been a chance to fix this, Andrey would have done this after all these years.

I disagree. Yes, it might be hard to solve the problems with the symbolic engine. I am fine with Andrey waiting for a better world to solve the problem.

You are direct Martin as always , and I think we discussed these things many times. I do not know what is Andrey waiting for, but it is lasting to long. As far I remember, I do not think that Andrey made any detailed analysis or explanation of all the things related to symbolic engine.

Wrote

But it cannot be that hard to set the default optimization to none (if I can change the setting, the core program should be able to do so as well).

Yes, you are right. It was proposed many time as a workaround. Of course, I do not think this would solve the problem.

Wrote

Also, the distinct operators for symbolic or numeric optimization should not provide prohibitive challenges. As you can see in the live version, there is a right arrow for symbolic evaluation and a = for numeric. This used to be this way in older versions of SMath and has been removed deliberately (Mathcad patent issues?). But again: The problem is in the optimization settings of assignments, not of the evaluation/display.

I guess that the symbolic engine is something Andrey has put into quite some effort and just cannot stand the idea of switching it off by default.

As I said, we can only guess about all of this.

Wrote

This reminds me of what is known as trap of emotional commitment in mechanical design. Designers tend to use 3D CAD models right from the start of a project. They focus on details, which is immediately rewarded by nice looking models. Questioning the basic concept or even the customer's requirements does not immediately pay but can avoid failure in the long run. The further you go in detailed design, the higher the costs of and the resistance to conceptual changes.

But perhaps it is not Andrey who stepped into the trap but me and Radovan and others. Once we went through the pain of finding out how to operate SMath, we tried to reward ourselves by sharing the word to others, collecting thanks in the forum. Do you remember, how long it took for you to find out how to write ranges a = 1..10? For me, that was several days. In order to not write off this as waste of time, I wrote the handbook. And so on. But doesn't the handbook contribute to the illusion (it is an illusion, given the recent bug report by Dijay) that SMath can be used for serious work? Isn't it a danger to the public if people start to use SMath for designing real things?

I still have hope that this would not be a contribution to an illusion and all of us here will not fill like losers. To be honest, in this case I felt like one many times. I would be very relieved if I could have seen the version SMath 1.0 and could have said - "That's it!".

Wrote

A program without an active user community is dead. In order to keep the community active, there must be some safe and reliable use cases. That is why we ask for default optimization settings being None or numeric. SMath can still be very useful without symbolic optimization of assignments.

"From your words to God's ear"

Regards,
Radovan

I tried to nail down the problem to a minimum working example. It is still not clear to me where the symbolic problem actually is but perhaps this inspires further investigations...

---

absroot.sm (4 КиБ) скачан 48 раз(а).

Wrote

I tried to nail down the problem to a minimum working example. It is still not clear to me where the symbolic problem actually is but perhaps this inspires further investigations...

A part of the issue: |0.9|

---

Working to fix it.

UPDATE: The issue not in abs(..) simplification.
UPDATE: Fixed.

Wrote

If I'm right and the author is interested in testing results I propose another way: Set the numeric Modus as Default, install a test button for symbolic modus and automatic comparison of the results and aks the community to help testing!

I strongly support this proposal. This, however, would require to modify the definitions one by one and evaluate all subsequent results. This could be done in the background (with an option to switch it off). Whenever a difference occures, then the user should be offered to submit a corresponding report. That could reveal much more bugs than we can find just by chance.

Still I strongly support SS-19 and SS-65.

How about a Math region utility tool? If it is possible to change the font of text regions, then it should be possible to change all math definitions to numeric.

Wrote

Fixed.

Many thanks Andrey

Wrote

But it cannot be that hard to set the default optimization to none (if I can change the setting, the core program should be able to do so as well).

Also I'm waiting for another default optimization (numeric or none); as I've written in a proposal/sub-proposal in the BTS, should be hardcoded or by choice (global settings)

Wrote

Also, the distinct operators for symbolic or numeric optimization should not provide prohibitive challenges. As you can see in the live version, there is a right arrow for symbolic evaluation and a = for numeric.

Just to introduce the taste in this small thing, I don't like very much the right arrow, I prefer something like a colored (blue?) equal an equal with a small colored s over or something else (printable by choice as for definitions); maybe could be a long step, but an advanced search tool that select math regions also by optimizations could be another way.

Wrote

Isn't it a danger to the public if people start to use SMath for designing real things?

Maybe, but not much more than using some design software or some large spreadsheet... as usual if you don't check the results in some different way, you can fall in trouble. And knowing weaknesses most of the issues are not so problematic

Wrote

If it is possible to change the font of text regions, then it should be possible to change all math definitions to numeric.

CTRL+A -> context menu choice over a single region already does the job

Wrote

Wrote

If it is possible to change the font of text regions, then it should be possible to change all math definitions to numeric.

CTRL+A -> context menu choice over a single region already does the job

This also changes the setting of all displays to numeric, which is not what I propose, because it prevents using symbolic engine at all. It's just about the definitions (assignments), because it's their symbolic setting which causes the trouble in most cases.

Many thanks to all, especially for quick Fixing the Problem. Will Keep testing.