plot2d not ignoring errors within functions

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:39 Created by macrakis on 2021-10-26 18:50:53 Original: https://sourceforge.net/p/maxima/bugs/3881

---

plot2d correctly ignores isolated numeric evaluation errors when the first argument is an expression, but when it is a function, the error stops evaluation and no plot is made.

ff(x):=gamma(x)$
plot2d(ff,[x,-1,1]) => Maxima encountered a Lisp error: gamma: gamma(-1.0) is undefined. (no plot)

plot2d(lambda([x],gamma(x)),[x,-1,1]) => Maxima encountered a Lisp error: gamma: gamma(-1.0) is undefined. (no plot)

but

plot2d(gamma(x),[x,-1,1]) => plot2d: expression evaluates to non-numeric value somewhere in plotting range. (but makes plot with the good points)

All three should function the same.

Strangely, if the function is 1/x, it works fine.

Maxima 5.45.1 / SBCL 2.0.0 / Windows 10

Also fails using the current version of plot.lisp on sourceforge (https://sourceforge.net/p/maxima/code/ci/1e9aaad02f86b4acce3cbd375a65499440ffabda/tree/src/plot.lisp)

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:41 Created by macrakis on 2021-10-26 18:51:58 Original: https://sourceforge.net/p/maxima/bugs/3881/#a86f

---

• assigned_to: Jaime E. Villate

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:44 Created by macrakis on 2021-10-26 18:54:13 Original: https://sourceforge.net/p/maxima/bugs/3881/#1bb4

---

• Description has changed:

Diff:

--- old
+++ new
@@ -12,6 +12,8 @@

 All three should function the same.

+Strangely, if the function is 1/x, it works fine.
+
 Maxima 5.45.1 / SBCL 2.0.0 / Windows 10

 Also fails using the current version of plot.lisp on sourceforge (https://sourceforge.net/p/maxima/code/ci/1e9aaad02f86b4acce3cbd375a65499440ffabda/tree/src/plot.lisp)

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:48 Created by villate on 2022-02-08 16:12:52 Original: https://sourceforge.net/p/maxima/bugs/3881/#6d43

---

This might have to do with the gamma function. Observe that the following 3 plot commands give the same result:

gg(x):=1/x+1/(x+1)$
plot2d(gg,[x,-1,1])$
plot2d(1/x+1/(x+1),[x,-1,1])$
plot2d(lambda([x],1/x+1/(x+1)),[x,-1,1])$

Notice also the following behavior of coerce-float-fun:

(funcall (coerce-float-fun #$gamma(x)$ #$[x]$) -1)
            -2.5653050788007548e16
(funcall (coerce-float-fun #$gamma(x)$ #$[x]$) -1.0)
             gamma: gamma(-1.0) is undefined.

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:51 Created by robert_dodier on 2022-03-02 07:46:12 Original: https://sourceforge.net/p/maxima/bugs/3881/#4d03

---

I don't think this behavior is specific to gamma. There is some logic in %COERCE-FLOAT-FUN to catch errors in the constructed function, but it doesn't catch errors generated by MERROR, which is how errors are triggered in many places in Maxima (and in SIMPGAMMA in particular).

I tried tinkering with it to try to get it to catch the error thrown by MERROR but couldn't get it to work. I'm probably overlooking something simple.

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:55 Created by kjak on 2022-03-09 20:32:11 Original: https://sourceforge.net/p/maxima/bugs/3881/#c739

---

I didn't get any response about my patch, so I'll commit something like it tomorrow if there are no objections.

One possible change -- which I'll include if I don't hear anything otherwise -- is binding errormsg to false around the ERRCATCH forms. My posted patch does not do this, and it causes the error messages to be printed before the "plot2d: expression evaluates to non-numeric value somewhere in plotting range." message.

Currently (without my patch) error messages aren't printed, so we have this:

(%i1) plot2d(lambda([x],log(abs(x))),[x,-1,1])$
plot2d: expression evaluates to non-numeric value somewhere in plotting range.

(%i2) plot2d(lambda([x],1/x),[x,-1,1])$
plot2d: expression evaluates to non-numeric value somewhere in plotting range.

But I actually prefer having the error messages printed, so something like this:

(%i4) plot2d(lambda([x],1/x),[x,-1,1])$
expt: undefined: 0 to a negative exponent.
plot2d: expression evaluates to non-numeric value somewhere in plotting range.

(%i5) plot2d(lambda([x],log(abs(x))),[x,-1,1])$
log: encountered log(0).
plot2d: expression evaluates to non-numeric value somewhere in plotting range.

(%i6) plot2d(lambda([x],gamma(x)),[x,-1,1])$
gamma: gamma(-1.0) is undefined.
gamma: gamma(0.0) is undefined.
plot2d: expression evaluates to non-numeric value somewhere in plotting range.

Presumably we want to keep the current behavior though.

[rtoy]

Imported from SourceForge on 2024-07-09 19:42:58 Created by peterpall on 2022-03-09 21:46:07 Original: https://sourceforge.net/p/maxima/bugs/3881/#aba6

---

Without the error messages it sometimes is hard to find out what went wrong causing the non-numeric values.

[rtoy]

Imported from SourceForge on 2024-07-09 19:43:02 Created by kjak on 2022-03-09 21:55:25 Original: https://sourceforge.net/p/maxima/bugs/3881/#aba6/495b

---

Yes, that's one of the reasons why I prefer to print the error messages.

[rtoy]

Imported from SourceForge on 2024-07-09 19:43:06 Created by villate on 2022-03-10 08:19:48 Original: https://sourceforge.net/p/maxima/bugs/3881/#d3c5

---

Kris: please go ahead and commit your patch. I actually prefer those extra error messages.