Threaded Mode | Linear Mode

HrastProgrammer · 06-03-2015, 05:33 AM

<HighLevelMode On>

I use both RPN and algebraic notation. When programming I always prefer algebraic notation (that's why I would take Pascal/BASIC/C/C++ over RPL anytime). For simple calculations RPN is probably the most convenient method but for complex calculations I appreciate the ability to recall the whole expression, make the necessary corrections and evaluate it again.

<HighLevelMode Off>

<LowLevelMode On>

Well, in assembler it's either a matter of direct calculations using registers, or some kind of stack machine (=RPN) anyway ...

<LowLevelMode Off>

Sukiari · 06-03-2015, 03:04 PM

(06-02-2015 02:40 AM)AnalogJoe Wrote:
(06-02-2015 01:27 AM)Andreas Grund Wrote: My wife and my daughters don't like RPN calculators, for them I am obsessed. ...

HAHA, I would love to hear what they say to you, I mean im single so nobody bugs me and I dont care, but I can only wonder what your "iphone generation" daughter thinks about his father with more than 20 weird old calculators.

My wife has been a math teacher for years now, first at the university level and now that we've moved into the hinterlands, in a middle school. Hardly anybody in the math field uses an RPN calculator but there were a few young and old who did at the university.

The iPhone generation generally uses Wolfram Alpha as their graphing calculator. While phones are not allowed in standardized tests, they are by and large tolerated in class especially as a learning aid. I suspect that most of the high schoolers who buy a graphing calculator these days will only use it during tests where their phone is disallowed. Non-RPN / RPL calculators make more sense for this application because there is only a very short learning curve.

Anyway I suspect the pocket calculator will always be with us as long as we maintain a civilization capable of producing electronics, but I also suspect that RPN will continue to be somewhat of a niche system, and RPL is now officially 'done for' with HP's plans to discontinue the 50g. I got my spare 50g - did YOU?

/sweeps finger across virtual room

Sukiari · 06-03-2015, 03:07 PM

(06-02-2015 01:59 PM)TASP Wrote: The 41CX em app will go on my next phone. I'm eligible for a new phone and am considering an Apple 5s. My Droid has been a big disappointment. It is a very early version, had some disturbing issues with apps the first month I had it, and subsequent software upgrades have further crippled it.

I haven't had a decent cellphone since my Chocolate.

I can recommend i41CX+ - while it isn't cheap it is full featured and amazing. How about pulling your GPS coordinates onto the stack or using the accelerometer in your programs? It also includes a recent version of Derive which interfaces with the stack, so you can use a modern CAS in your RPN programs!

Seriously awesome bit of software there.

AnalogJoe · (This post was last modified: 06-04-2015 12:55 AM by AnalogJoe.)

(06-03-2015 03:04 PM)Sukiari Wrote: My wife has been a math teacher for years now, first at the university level and now that we've moved into the hinterlands, in a middle school. Hardly anybody in the math field uses an RPN calculator but there were a few young and old who did at the university.

The iPhone generation generally uses Wolfram Alpha as their graphing calculator. While phones are not allowed in standardized tests, they are by and large tolerated in class especially as a learning aid. I suspect that most of the high schoolers who buy a graphing calculator these days will only use it during tests where their phone is disallowed. Non-RPN / RPL calculators make more sense for this application because there is only a very short learning curve.

Anyway I suspect the pocket calculator will always be with us as long as we maintain a civilization capable of producing electronics, but I also suspect that RPN will continue to be somewhat of a niche system, and RPL is now officially 'done for' with HP's plans to discontinue the 50g. I got my spare 50g - did YOU?

/sweeps finger across virtual room

Wow, I cant understand how someone like your wife who's a math teach doesnt appreciate the value of RPN. But yes, Im an engineer and most of my colleagues love TI calculators, most of the heavy work is done with matlab, but calculators are still the prefered choice for quick calculations. Some things are just faster with calculators, for example, I've timed myself (yes im a nerd) and for stuff like solving linear systems of equations it takes more time for me to make a matrix array in matlab, and perform the necessary operations to get the answer, than using for instance the Linsolver tool on the HP50G.

Crunching numbers is also by far faster and easier on a calculator than a computer, im always like "where the heck is the ^ simbol on the keyboard" or stuff like that, also its a lot easier to just hit the key "sin" or "cos" than typing an entire command like sin(4).

And people who suggest iPhone apps. Im sorry I dont own a smartphone, so I couldnt tell...., but Ill just say this: I need REAL keys.

Im glad that RPN is still an option on the Prime, people argue that the Prime is not fully RPN like in the way its programmed or in the CAS window, well yeah but it is RPN where it matters the most which is in the HOME window where you perform all the number crunching, and thats fine by me...

As for the HP 50G, yeap I have my spare, at one time I thought that I should invest on several HP 48GX but Ive learned to replace the 48G with the 50G and I could definitely live without the 48GX. The 50G is like a fast 48GX with a cheaper keyboard and some new features, but overall they are both the same thing.

I hope HP keeps offering the RPN option in their future calculators, otherwise the Prime will be my last HP calc.

Marcio · (This post was last modified: 06-04-2015 01:12 AM by Marcio.)

(06-04-2015 12:54 AM)AnalogJoe Wrote: I hope HP keeps offering the RPN option in their future calculators, otherwise the Prime will be my last HP calc.

Maybe they will release the 36s this year. An improved 35s with all bugs fixed and the elementary stuff that could have been easily implemented but were not. Also, complex numbers for those who need it.

And the most important thing of all: Build it at home!

Tugdual · (This post was last modified: 06-04-2015 05:58 AM by Tugdual.)

For a very long time I've been using a 15C and nothing else. I actually did my high studies with a 15C and I totally knew all features (except Brouhaha's synthetics Big Grin

).
RPN became a second nature and with the excellent keyboard I was extremely fast (reason why I don't believe in phones to replace calculators).
A few years ago I decided to have a look on more recent features such as CAS and I bought a Prime. Was pretty disappointed for many reasons: RPN had almost disappeared, CAS was not impressive and Basic totally inefficient for a handled machine (Prime resembles to a computer, except that computers are much better than a Prime). I then had a 15CLE when it was sold for an acceptable price when it was released. Nothing new of course apart from speed and quick battery depletion.
I moved on 35s as a cheap every day calculator and, to my surprise, I found it as good as the 15C, some aspects even better but here again, no dramatic change in experience.
I finally acquired a 48G that I really found very exciting but slow and screen difficult to read. I then discovered that the 50g was pretty much the same as 48G but a bit better (despite plastic case and keyboard). This is now the machine I use the more. The RPL is in my opinion totally aligned with RPN concepts and handled machine coding approach. Better than keystrokes coding much more adapted than basic. I also love the Fx keys, the VAR, the MES solver, easy (compared to the Prime) unit of measurement, easy (same comment) base, the infinite stack. I don't care much about graphics and I stick on userRPL, I'm not a big fan of cryptic coding, calculator has to be simple and straightforward. This machine is a monster not that difficult to tame and very good for a large spectrum of operations, from simple to complex. The 50g is definitely a breakthrough after the 15C and 35s. BTW for fun, I took my original 15C user guide and started the exercises with the 50G. So easy... that is an interesting way to measure progress.

Dave Britten · 06-04-2015, 01:16 PM

(06-03-2015 05:33 AM)HrastProgrammer Wrote: <HighLevelMode On>

I use both RPN and algebraic notation. When programming I always prefer algebraic notation (that's why I would take Pascal/BASIC/C/C++ over RPL anytime). For simple calculations RPN is probably the most convenient method but for complex calculations I appreciate the ability to recall the whole expression, make the necessary corrections and evaluate it again.

<HighLevelMode Off>

<LowLevelMode On>

Well, in assembler it's either a matter of direct calculations using registers, or some kind of stack machine (=RPN) anyway ...

<LowLevelMode Off>

I have to confess to agreeing with you. RPN (and RPL) is fantastic for number crunching, or small programs that augment the built-in functions, but for building a large, sophisticated application, it's painful at best, especially with no GOTO in RPL. Compared to a symbolic language, the code is almost always opaque and barely maintainable, with no immediately obvious connection to what's getting passed as a parameter to what, and a lack of any code commenting ability. (RPL is often called a write-only language, and I think I've heard the same epithet applied to perl and/or regular expressions.)

If I'm writing self-contained math routines, or small utility functions, an RPN/RPL machine is great for that job, and usually my first choice. But if I'm writing a program for data logging, estimating process completion via a choice of regression models on a user-selectable most-recent subset of the data, and optionally plotting the logged data points against the regression line*? Hand me the TI-89.

*I am actually doing this right now. The 89 does make me miss the alarm/audio functionality of the 48, though, that's for sure.

Marcio · 06-04-2015, 01:46 PM

(06-04-2015 01:16 PM)Dave Britten Wrote: If I'm writing self-contained math routines, or small utility functions, an RPN/RPL machine is great for that job, and usually my first choice. But if I'm writing a program for data logging, estimating process completion via a choice of regression models on a user-selectable most-recent subset of the data, and optionally plotting the logged data points against the regression line*? Hand me the TI-89.

I think the Prime can do that too, might be even faster.

Dave Britten · 06-04-2015, 03:07 PM

(06-04-2015 01:46 PM)Marcio Wrote:
(06-04-2015 01:16 PM)Dave Britten Wrote: If I'm writing self-contained math routines, or small utility functions, an RPN/RPL machine is great for that job, and usually my first choice. But if I'm writing a program for data logging, estimating process completion via a choice of regression models on a user-selectable most-recent subset of the data, and optionally plotting the logged data points against the regression line*? Hand me the TI-89.

I think the Prime can do that too, might be even faster.

Most likely. But based on my occasional excursions to the Prime forum, it also seems to be full of assorted bugs and other annoyances still.

Tugdual · 06-04-2015, 03:54 PM

(06-04-2015 01:16 PM)Dave Britten Wrote:
(06-03-2015 05:33 AM)HrastProgrammer Wrote: <HighLevelMode On>

I use both RPN and algebraic notation. When programming I always prefer algebraic notation (that's why I would take Pascal/BASIC/C/C++ over RPL anytime). For simple calculations RPN is probably the most convenient method but for complex calculations I appreciate the ability to recall the whole expression, make the necessary corrections and evaluate it again.

<HighLevelMode Off>

<LowLevelMode On>

Well, in assembler it's either a matter of direct calculations using registers, or some kind of stack machine (=RPN) anyway ...

<LowLevelMode Off>

I have to confess to agreeing with you. RPN (and RPL) is fantastic for number crunching, or small programs that augment the built-in functions, but for building a large, sophisticated application, it's painful at best, especially with no GOTO in RPL. Compared to a symbolic language, the code is almost always opaque and barely maintainable, with no immediately obvious connection to what's getting passed as a parameter to what, and a lack of any code commenting ability. (RPL is often called a write-only language, and I think I've heard the same epithet applied to perl and/or regular expressions.)

If I'm writing self-contained math routines, or small utility functions, an RPN/RPL machine is great for that job, and usually my first choice. But if I'm writing a program for data logging, estimating process completion via a choice of regression models on a user-selectable most-recent subset of the data, and optionally plotting the logged data points against the regression line*? Hand me the TI-89.

*I am actually doing this right now. The 89 does make me miss the alarm/audio functionality of the 48, though, that's for sure.

I personally tend to avoid complex programs on calculators, for the reason that keyboard is not sized for this activity; so RPL seems the perfect answer to me. Now to your point, you often need multiple programs to achieve one task for the reason you mentioned which again shows that the calculator is not the right tool for bigger applications.

Dave Britten · 06-04-2015, 07:24 PM

(06-04-2015 03:54 PM)Tugdual Wrote: I personally tend to avoid complex programs on calculators, for the reason that keyboard is not sized for this activity; so RPL seems the perfect answer to me. Now to your point, you often need multiple programs to achieve one task for the reason you mentioned which again shows that the calculator is not the right tool for bigger applications.

A complex program doesn't necessarily have to have a complex user interface, though.

The ETA program I'm finishing up displays a menu bar when you run it. Just hit F1 ("Log") and enter the current value to log it to the data lists, along with the current timestamp (retrieved automatically from the RTC). If you have enough data points, it then runs the regression automatically, and displays the completion estimate and correlation. Then you can use the "Plot" menu item to draw the graph. There's also a "Setup" option where you can enter the target value, select a regression model, and choose how many of the most recent data points to look at when estimating (or 0 to use all of them).

It's a relatively large program for a calculator, but the user interface has been kept simple and similar to the other built-in apps. I've written a similar program on my 48SX. It's much more basic overall - it simply expects you to store a value in TARG, then enter a data value on the stack and run ETA to log a data point. Then it runs the recalculation routine, and there's another program that will draw a plot if you want. Just with those features, the code is still a lot uglier and unpleasant to maintain.

The RPN programming paradigm tends to shine more with the kind of stuff you'd do on a 42S/41C/32S/etc. The TI-89 is a lot clumsier for interactive use due to the symbolic command entry, but programming larger applications is much more straight-forward.

Thomas Klemm · 06-05-2015, 11:54 PM

(06-04-2015 01:16 PM)Dave Britten Wrote: RPN (and RPL) is fantastic for number crunching, or small programs that augment the built-in functions, but for building a large, sophisticated application, it's painful at best, especially with no GOTO in RPL. Compared to a symbolic language, the code is almost always opaque and barely maintainable, with no immediately obvious connection to what's getting passed as a parameter to what, and a lack of any code commenting ability.

Solving a quadratic equation is probably something all of us are familiar with:

\( ax^2 + bx + c = 0 \)

However I'm not using the quadratic formula but transform the equation into this reduced form:

\( x^2 - 2px + q = 0 \)

By comparing the coefficients we find:

\( p = \frac{b}{-2a} \)

\( q = \frac{c}{a} \)

The solutions for this equation are:

\( x_{1,2} = p \pm D \)

where \( D = \sqrt{p^2-q} \)

Compare these two solutions:

Code:

\<<

    \-> a b c \<<

        b a / -2 /

        c a /

        \-> p q \<<

            p SQ q - \v/

            \-> D \<<

                p D -

                p D +

            \>>

        \>>

    \>>

\>>

Code:

\<<

    ROT ROT OVER / -2 /

    ROT ROT /

    OVER SQ SWAP - \v/

    DUP2 - 

    ROT ROT +

\>>

Which one is easier to understand?

But we can split the calculation into three tasks:

TRANSFORM ( a b c -- p q )
DISCRIMINANT ( p q -- D )
PLUSMINUS ( p D -- x₁ x₂ )

Code:

%%HP: T(3)A(R)F(.);

DIR

  SOLVE \<<

        TRANSFORM

        DISCRIMINANT

        PLUSMINUS

  \>>

  UNROT \<<

        ROT ROT

  \>>

  TRANSFORM \<<

        UNROT OVER / -2 /

        UNROT /

  \>>

  DISCRIMINANT \<<

        OVER SQ SWAP - \v/

  \>>

  PLUSMINUS \<<

        DUP2 -

        UNROT +

  \>>

END

This also allows to test each of them individually.

Quote:(RPL is often called a write-only language, and I think I've heard the same epithet applied to perl and/or regular expressions.)

Whether your code is maintainable has little to do with the programming language you use.
It is more related to how the code is organized.
Using subroutines or local variables is a way to give meaningful names. This helps to understand and thus maintain the code.

Cheers
Thomas

Dave Britten · 06-06-2015, 03:07 AM

(06-05-2015 11:54 PM)Thomas Klemm Wrote: But we can split the calculation into three tasks:

TRANSFORM ( a b c -- p q )

DISCRIMINANT ( p q -- D )

PLUSMINUS ( p D -- x₁ x₂ )

Code:

%%HP: T(3)A(R)F(.); DIR SOLVE \<< TRANSFORM DISCRIMINANT PLUSMINUS \>> UNROT \<< ROT ROT \>> TRANSFORM \<< UNROT OVER / -2 / UNROT / \>> DISCRIMINANT \<< OVER SQ SWAP - \v/ \>> PLUSMINUS \<< DUP2 - UNROT + \>> END

This also allows to test each of them individually.

That's the kind of approach I try to take on the 48, and to some extent on RPN models where appropriate/feasible, e.g. I made programs called XMIN, XMAX, etc, rather than opaque blobs of code pulling things out of PPAR.

(06-05-2015 11:54 PM)Thomas Klemm Wrote: Whether your code is maintainable has little to do with the programming language you use.
It is more related to how the code is organized.
Using subroutines or local variables is a way to give meaningful names. This helps to understand and thus maintain the code.

Cheers
Thomas

When the language lacks meaningful whitespace and inline comments, and functions can be getting their arguments from heaven knows where previously in the code, I wouldn't put all of the blame on the user.

Something like this lends itself even better to easy understanding: (\@ = weird TI-89 comment symbol)

Code:

quadrdce(a,b,c)

Func

\@Reduced quadratic equation:

\@x^2-2px+q=0

Local p,q,d

b/(-2*a)\->p

c/a\->q

\SQRT(p^2-q)\->d

Return {p+d,p-d}

EndFunc

So even though the 89 is really obnoxious to use as a number cruncher, it does offer a nice programming environment. The Prime has similar paradigms from the little that I've seen.

Les Bell · (This post was last modified: 06-06-2015 03:22 AM by Les Bell.)

I've posted a few times recently in discussions about RPN, particularly about the use of RPN to evaluate algebraic expressions from left to right.

I take the view that the RPN user's knowledge of the problem domain guides his approach to the problem and essentially avoids many ambiguities or pitfalls that seem to occur, and I've cooked up a very simple example to demonstrate.

If asked to calculate the surface area of a cylinder of 4m diameter (2m radius) and height 6m, I suspect many students - since this is the topic they are studying at the time - will have an algebraic equation in front of them and may already have reduced it to this form:

\[
A = 2 \pi r ( r + h)
\]
I suppose some RPN novices might plod from left to right along this: 2 PI * 2 * 2 ENTER^ 8 + *. The experienced RPN user will solve this from the inside out, with the keystrokes 2 ENTER^ ENTER^ 6 + * PI * 2 * (noticing that two copies of r (2) are required on the stack). This saves double-entering the radius

However, that's the mathematical approach to the problem.

The classic engineering RPN user will solve it by reasoning thusly:

"First I need the area of the end cap - that's \(\pi r^2\), so I enter 2 ENTER^ * PI *, and there's two of those, so 2 *. Now I need the area of the vertical wall, which will be the circumference times the height, so that's 4 ENTER^ PI * 6 *. And I need to add that to the area of the ends, which is already on the stack, so +."

All up, you get: 2 ENTER^ * PI * 2 * 4 ENTER^ PI * 6 * +. It's 4 more keystrokes than the mathematical approach, but it's done entirely on the fly, with no paper-and-pencil work or reference to formulas, other than the most basic.

RPN is great for supporting that kind of reasoning - you work towards the answer via a series of intermediate steps, which are retained on the stack as you go.

(06-05-2015 11:54 PM)Thomas Klemm Wrote: Whether your code is maintainable has little to do with the programming language you use.
It is more related to how the code is organized.
Using subroutines or local variables is a way to give meaningful names. This helps to understand and thus maintain the code.

RPL programming is a bit like that, as Thomas's post shows; it seems to me that you're meant build up a complete RPL program by composing it from lower-level functions. It would be easy enough to do an RPL version of my example by writing RPL functions for the area of a circle, etc. and composing them into an RPL program.

Tugdual · 06-06-2015, 07:37 AM

I would like to mention that the '\' character is not visible on the 50g. This is only an artefact when publishing the source code and I believe this is only intended to transfer the code from machine to machine. I find it unfair to publish source code with noisy characters and qualify it difficult to read and understand. This is somehow equivalent to reading HTML tags instead of viewing the final page.

Thomas Klemm · 06-06-2015, 08:42 AM

(06-06-2015 07:37 AM)Tugdual Wrote: I find it unfair to publish source code with noisy characters and qualify it difficult to read and understand.

This was certainly not my intent.
Hopefully this is better:

Code:

« → a b c

  « b a / -2 /

    c a / → p q

    « p SQ q - √ → D

      « p D -

        p D +

      »

    »

  »

»

Code:

«

  ROT ROT OVER / -2 /

  ROT ROT /

  OVER SQ SWAP - √

  DUP2 - 

  ROT ROT +

»

Cheers
Thomas

Tugdual · 06-06-2015, 09:33 AM

(06-06-2015 08:42 AM)Thomas Klemm Wrote:
(06-06-2015 07:37 AM)Tugdual Wrote: I find it unfair to publish source code with noisy characters and qualify it difficult to read and understand.

This was certainly not my intent.
Hopefully this is better:

Code:

« → a b c « b a / -2 / c a / → p q « p SQ q - √ → D « p D - p D + » » » »

Code:

« ROT ROT OVER / -2 / ROT ROT / OVER SQ SWAP - √ DUP2 - ROT ROT + »

Cheers
Thomas

Nothing personal Thomas, I know there are new readers here wondering if the 50g is the right choice and I just wanted to clarify this aspect. I often struggle with the tone of my messages...

Dave Britten · 06-06-2015, 12:11 PM

(06-06-2015 07:37 AM)Tugdual Wrote: I would like to mention that the '\' character is not visible on the 50g. This is only an artefact when publishing the source code and I believe this is only intended to transfer the code from machine to machine. I find it unfair to publish source code with noisy characters and qualify it difficult to read and understand. This is somehow equivalent to reading HTML tags instead of viewing the final page.

Yeah, I think most of us are just accustomed to typing the code the way translate 3 would render it, strange as it may look at times.

But reading HTML tags is certainly fair game if you're doing any degree of web development.

Sukiari · 06-10-2015, 09:31 PM

(06-04-2015 12:54 AM)AnalogJoe Wrote: Wow, I cant understand how someone like your wife who's a math teach doesnt appreciate the value of RPN. But yes, Im an engineer and most of my colleagues love TI calculators

It's probably the same thing with QWERTY and Dvorak keyboards - why does anybody use QWERTY? Because that's what was put in front of you when you learned to type. Personally I find either input method to be workable but RPN seems to be my preference. My first "real" calculator was my old Apple IIGS, then eventually a TI-85 was my first scientific / graphing calc. Both are infix.

(06-04-2015 12:54 AM)AnalogJoe Wrote: Im glad that RPN is still an option on the Prime, people argue that the Prime is not fully RPN like in the way its programmed or in the CAS window, well yeah but it is RPN where it matters the most which is in the HOME window where you perform all the number crunching, and thats fine by me...

I think people have many legitimate complaints about the Prime, and I hope HP listens to them. While it's not for me (it its current form) some day I might get one.

(06-04-2015 12:54 AM)AnalogJoe Wrote: As for the HP 50G, yeap I have my spare, at one time I thought that I should invest on several HP 48GX but Ive learned to replace the 48G with the 50G and I could definitely live without the 48GX. The 50G is like a fast 48GX with a cheaper keyboard and some new features, but overall they are both the same thing.

I hope HP keeps offering the RPN option in their future calculators, otherwise the Prime will be my last HP calc.

I agree with you about the 50G's place in the HP lineup. It's a great machine, perhaps the last vestige of the old HP. I don't really even think it's that much cheaper than the 48 series, which are creaky and have the critical flaw that connections work loose. It *LOOKS* cheaper than the 48 series though, that's for sure.

AnalogJoe · 06-19-2015, 01:03 AM

(06-10-2015 09:31 PM)Sukiari Wrote: I think people have many legitimate complaints about the Prime, and I hope HP listens to them. While it's not for me (it its current form) some day I might get one.

I agree with you about the 50G's place in the HP lineup. It's a great machine, perhaps the last vestige of the old HP. I don't really even think it's that much cheaper than the 48 series, which are creaky and have the critical flaw that connections work loose. It *LOOKS* cheaper than the 48 series though, that's for sure.

Well, Ive definitely read some horror stories about the Prime, however I dont know if im just lucky or what but my Prime works perfectly.

Ill tell you something thou, I have two HP 50Gs, one of them is older and the other is quite recent, and the newer one does seem considerably "cheaper" than the older one, to begin with, the older one came in a leather or leather like pouch, while the newer one came in a cheap cloth pouch. The plastic on the new one seems a lot more brittle and thin, the older one feels a lot stronger, and the plastic is a lot better looking, the new one is matte, while the old one is black with like shinny sprinkles. Even the size is slightly different the newer one is a bit longer, while the older one is shorter but a bit wider, both have around the same thickness. Even the HP logo on the older one seems a lot better quality.

Recently Ive seen the 50G dropping price considerably, most stores Ive visited are offering the 50G on a discount and many stores are selling the last of them, does anyone knows if there are any plans on discontinuing the 50G in the near future?