List Commands Library for 50g
01-28-2019, 11:47 PM
Post: #401
 Thomas Klemm Senior Member Posts: 1,447 Joined: Dec 2013
RE: List Commands Library for 50g
(01-27-2019 10:02 PM)Thomas Okken Wrote:  The probability of RAND returning zero is exactly zero. It can't happen.

Interestingly enough, this is not the case with neither the HP-11C nor the HP-15C:
(10-21-2018 09:03 PM)Thomas Klemm Wrote:  Just try this to get 0:

8.603685347E-1
STO RAN#
RAN#
01-29-2019, 12:16 AM
Post: #402
 ttw Member Posts: 219 Joined: Jun 2014
RE: List Commands Library for 50g
A quick method to avoid zero (and one which also gives problems) in a pseudo random number generator is to generate uniformly distributed integers over (0,M-1) or (1,M) for some large (prime or power of two or whatever) integer. Then one computed for result X (0<=X<M) the number (2X+1)/2N as either a fraction or a floating point number.
01-29-2019, 06:20 PM
Post: #403
 John Keith Senior Member Posts: 643 Joined: Dec 2013
RE: List Commands Library for 50g
(01-27-2019 07:07 PM)pier4r Wrote:  About RAND a * b + IP. I used it too at first but it messes up when integers are negative if I am not mistaken.

Sure the mode that works also with arbitrary integers is neat. If you fix it I put it in the collection too. Could you ping me if you do it?

Program has been fixed. Replacing IP with FLOOR is all it took.
01-29-2019, 07:00 PM (This post was last modified: 01-29-2019 07:01 PM by pier4r.)
Post: #404
 pier4r Senior Member Posts: 2,067 Joined: Nov 2014
RE: List Commands Library for 50g
(01-27-2019 06:52 PM)John Keith Wrote:  Edited to fix bug that would cause negative values to be off by 1.

Code:
 \<< 3. \->LIST I\->R EVAL OVER - 1. + \-> s b a   \<<     IF -105. FS?     THEN 1. s       START RAND a * b + FLOOR       NEXT     ELSE 1. s       START RAND a * b + FLOOR R\->I       NEXT     END s \->LIST   \>> \>>

I am still not convinced. But thanks for the effort!

I mean say we have a range of: -5 to 5. When you do RAND a * b + you pick a value, actually, between 0 - eps and -4 - eps (where eps<1). Then you add 5 to it, so one gets mostly positive integers or zero.

Am I missing something obvious here?

Wikis are great, Contribute :)
02-01-2019, 08:14 PM
Post: #405
 John Keith Senior Member Posts: 643 Joined: Dec 2013
RE: List Commands Library for 50g
(01-29-2019 07:00 PM)pier4r Wrote:  I mean say we have a range of: -5 to 5. When you do RAND a * b + you pick a value, actually, between 0 - eps and -4 - eps (where eps<1). Then you add 5 to it, so one gets mostly positive integers or zero.

Am I missing something obvious here?

I don't know, but I just ran the revised program and got a list containing approximately equal numbers of each of the integers from -5 through 5.

Try it yourself: run the program with the arguments 1000, -5, 5 then execute LSORT LRPCT on the resulting list to see the stats.
02-01-2019, 10:42 PM (This post was last modified: 02-01-2019 10:53 PM by pier4r.)
Post: #406
 pier4r Senior Member Posts: 2,067 Joined: Nov 2014
RE: List Commands Library for 50g
Ok went debugging your program. I missed some points, not easy to read in the code at glance.
Your 'a' is not the start of the range (intuitively: a is the start and b is the end).
Rather a is the size of the range, b is the start of it.

Then it is sound yes.

Wikis are great, Contribute :)
05-13-2019, 11:36 AM
Post: #407
 Gilles Member Posts: 171 Joined: Oct 2014
RE: List Commands Library for 50g
Hi,

if I found some time, i will try to create a ListExt Library (or a subset) in NewRPL. But I'm afraid about the DOPERM and DOCOMB commands... DavidM, if you read this, what algorithm did you use?
05-13-2019, 03:43 PM (This post was last modified: 05-17-2019 02:57 PM by DavidM.)
Post: #408
 DavidM Senior Member Posts: 789 Joined: Dec 2013
RE: List Commands Library for 50g
I ended up spending a lot of time on those, starting down several different algorithmic paths and abandoning ideas at various stages for one reason or another. DOCOMB can be summarized as follows:

Code:
create a "mask" that represents the first group of list elements for each unique mask:    split the source list into two lists based on the current mask:       "selected" list elements       "remnant" list elements    execute the supplied user program with the new lists defined above    determine next lexicographic iteration of the mask group all results from the above into a final list

DOPERM is similar, but has an inner loop that steps through each permutation of each combination that is found above.

The mask is implemented as a string, with each "character" actually representing a zero-based index of a list element (0-255) for permutations and 0 or 1 for combinations.

In order for this approach to perform reasonably, the routines dealing with the mask are all coded in Saturn assembly.

There are some RPLish quirks that also had to be dealt with, especially for the user-supplied executable. It has to be "prepped" to convert any references to the special variables (XPRM, XCMB, CRMNT) into locals, since they will default to being globals when the user creates the program.
05-17-2019, 02:53 PM (This post was last modified: 05-17-2019 06:03 PM by DavidM.)
Post: #409
 DavidM Senior Member Posts: 789 Joined: Dec 2013
RE: List Commands Library for 50g
Here's a couple illustrative examples of how the masks I described above are used. I'm representing the masks in these examples as a string of numbers separated by spaces for clarity.

Example: a list of 5 items, we want all combinations of 3

Lexicographic stepping for DOCOMB/DOPERM combinations starts with the highest order form and works its way down to the lowest. So the first mask is created with a series of 1s that matches the count of selected elements, followed by 0s for the remainder:

1 1 1 0 0

The mask is then stepped down lexicographically in each iteration of the loop until the lowest order form is found:

1 1 1 0 0
1 1 0 1 0
1 1 0 0 1
...
0 1 1 0 1
0 1 0 1 1
0 0 1 1 1

The mask for each iteration is applied to the source list such that the resulting selection contains only the elements that correspond to the 1s. So if our source list is:

{ A B C D E }

...then the resulting combinations that correspond to the iterative masks are:

Code:
Src List/Mask  Resulting Sublist -------------  ----------------- A B C D E    1 1 1 0 0      A B C     1 1 0 1 0      A B   D   1 1 0 0 1      A B     E  1 0 1 1 0      A   C D  1 0 1 0 1      A   C   E  1 0 0 1 1      A     D E  0 1 1 1 0        B C D  0 1 1 0 1        B C   E  0 1 0 1 1        B   D E  0 0 1 1 1          C D E

Example: a list of 4 items

The first mask is generated with each index matching its zero-based starting position, and is thus already in the lowest-order form:

0 1 2 3

The mask is then stepped lexicographically upward until the highest order permutation is encountered:

0 1 2 3
0 1 3 2
0 2 1 3
...
3 1 2 0
3 2 0 1
3 2 1 0

The mask is applied to the source list in a similar fashion to the above description for combinations, but instead of a binary disposition (include/exclude), the mask designates the final positions of each of the source elements in the new list. So if our source list is:

{ A B C D }

...then the resulting permutation lists that correspond to the designated masks are as follows:

Code:
Src List/Mask  Resulting Permutation -------------  --------------------- A B C D 0 1 2 3        A B C D  0 1 3 2        A B D C  0 2 1 3        A C B D  0 2 3 1        A C D B  0 3 1 2        A D B C  0 3 2 1        A D C B  1 0 2 3        B A C D  1 0 3 2        B A D C  1 2 0 3        B C A D  1 2 3 0        B C D A  1 3 0 2        B D A C  1 3 2 0        B D C A  2 0 1 3        C A B D  2 0 3 1        C A D B  2 1 0 3        C B A D  2 1 3 0        C B D A  2 3 0 1        C D A B  2 3 1 0        C D B A  3 0 1 2        D A B C  3 0 2 1        D A C B  3 1 0 2        D B A C  3 1 2 0        D B C A  3 2 0 1        D C A B  3 2 1 0        D C B A

Stepping algorithm

The heart of all of this is a subroutine I called "NextMask", which takes a mask and a boolean indicating up/down stepping and returns the next mask and a boolean indicating if the new mask is at the high/low extreme. The algorithm is a modified form of this one, with the modifications allowing for up or down stepping. No changes were needed to accommodate combinations and permutations; the exact same algorithm inherently works for both so long as they are formatted as described above.

Hope this helps!
05-17-2019, 09:43 PM
Post: #410
 Gilles Member Posts: 171 Joined: Oct 2014
RE: List Commands Library for 50g
(05-17-2019 02:53 PM)DavidM Wrote:  Here's a couple illustrative examples of how the masks I described above are used. I'm representing the masks in these examples as a string of numbers separated by spaces for clarity.
Hope this helps!

Yes. Thank you very much for those explanations.
01-18-2021, 06:37 PM
Post: #411
 John Keith Senior Member Posts: 643 Joined: Dec 2013
RE: List Commands Library for 50g
Version 1.3 has been released: ListExt 1.3

This is a major update with many new commands.
01-18-2021, 10:03 PM
Post: #412
 DavidM Senior Member Posts: 789 Joined: Dec 2013
RE: List Commands Library for 50g
(01-18-2021 06:37 PM)John Keith Wrote:  Version 1.3 has been released: ListExt 1.3

This is a major update with many new commands.

Here's a quick rundown of the new commands. Many of the new commands operate identically to GoferLists commands with similar names, but are included in ListExt due to improved performance.

Short descriptions of the new commands are listed below, along with the general release notes. As always, complete descriptions and more examples are provided in the documentation that comes with the library.

Questions, comments, etc. are always welcome!

- David

Code:
LSCAN - returns a list of results from applying a program with each successive list element given    { 1 2 3 4 5 } « + » LSCAN => { 1 3 6 10 15 }    { 2 2 2 2 2 } « * » LSCAN => { 2 4 8 16 32 }      LMAP - maps a command or function to each list element (non-recursive)    { 1 2 3 4 5 } « 3 * » LMAP => { 3 6 9 12 15 }    { { 1 2 3 } { 1 2 } { 1 } } ::SIZE LMAP => { 3. 2. 1. } LXEQ - executes a program repeatedly with each list element as its argument    { 1 2 3 4 5 } « 3 * » LXEQ => 3 6 9 12 15       (similar to LMAP, but results are left on the stack instead of grouped into a list)     { "one" "two" "three" "four" } 4 { NXEQ DISP } CLLCD LXEQ 7 FREEZE => <no stack results>       (each object is displayed on a separate line starting on line 4 of the display) LDIFF - returns a list of the elements in list1 that aren't in list2    { 1 2 3 4 5 } { 2 3 4 5 6 } LDIFF => { 1 }    { 2 3 4 5 6 } { 1 2 3 4 5 } LDIFF => { 6 } LINTR - returns a list of the elements that are in both list1 and list2    { 1 2 3 } { 3 4 5 } LINTR => { 3 }    { 1 2 3 } { 5 4 3 2 1 } LINTR => { 1 2 3 } LSUBS - returns a list of all possible subgroupings of the elements in the given list    { 1 2 } LSUBS => { { } { 1 } { 2 } { 1 2 } }    { 1 2 3 } LSUBS => { { } { 1 } { 2 } { 1 2 } { 3 } { 1 3 } { 2 3 } { 1 2 3 } } LFILT - returns a list containing the elements for which the supplied program is TRUE    { 1 2 3 4 5 } « 2 > » LFILT => { 3 4 5 }    { 1 2 3 4 5 } { 2 OVER > SWAP 4 > OR } LFILT => { 1 5 } LWHL - executes a program while a given condition is TRUE, grouping the results into a list    2 ::NEXTPRIME { 30 < } LWHL => { 2 3 5 7 11 13 17 19 23 29 }    3 { DUP 2 MOD { 3 * 1 + } { 2 / } IFTE } { 1 SAME NOT } LWHL 1 + => { 3 10 5 16 8 4 2 1 } LDWHL - removes elements from the beginning of a list that match the provided condition    { 1 2 3 4 5 } « 4 < » LDWHL => { 4 5 }    { 1 2 3 4 5 } { 6 < } LDWHL => { } LTWHL - returns a list containing the first elements of a list that match the provided condition    { 1 2 3 4 5 } « 4 < » LTWHL => { 1 2 3 }    { 1 2 3 4 5 } { 6 < } LTWHL => { 1 2 3 4 5 } LCPRD - returns the Cartesian Product of two or more lists    { { 1 2 3 } { a b } } LCPRD => { { 1 a } { 1 b } { 2 a } { 2 b } { 3 a } { 3 b } }    { { 1 } { a b } { X Y Z } } LCPRD => { { 1 a X } { 1 a Y } { 1 a Z } { 1 b X } { 1 b Y } { 1 b Z } } LTAKE - returns a list containing the first or last elements specified by a count    { 1 2 3 4 5 } 3 LTAKE => { 1 2 3 }    { 1 2 3 4 5 } -3 LTAKE => { 3 4 5 } LDROP - returns a list containing all but the first or last elements specified by a count    { 1 2 3 4 5 } 3 LDROP => { 4 5 }    { 1 2 3 4 5 } -3 LDROP => { 1 2 } LCOMP - returns a list indicating matching and non-matching elements in 2 lists    { 1 2 3 4 5 } { 5 4 3 2 1 } 1 LCOMP => { 0. 0. 1. 0. 0. }    { 1 2 3 4 5 } { 5 4 3 2 1 } 0 LCOMP => { 1. 1. 0. 1. 1. } LDLTA - faster ΔLIST that also accepts lists with 0 or 1 element    { 1 } LDLTA => { 1 }    { 5 4 3 2 1 } LDLTA => { -1 -1 -1 -1 } NL→I2 - converts a list of numbers to an integer (faster, more specific input than NL→I)    { 1 2 3 } NL→I2 => 123    { 10 .00299 399 .04E-6 } NL→I2 => 1234

Here's the release notes for this version:

Version 1.3.0
January 4, 2021

- Refactored LRMOV to create a subroutine that could be used by other commands
- Added checks to the following commands for triggering garbage collection in low memory situations:
I→NL
NL→I
- Small speed improvement for SLST→
- Small speed improvement for LPICK
- Negative arguments for LDIST and LSDIV are now allowed; negative indicates quantity of sublists instead of size of each sublist
- LSHUF changes
- optimized to reduce initial list traversal
- reduced subroutine calls during shuffling
- SplitMix64 seed now incremented as per standard definition as opposed to using final output of each iteration as the new seed
- DOCOMB and DOPERM now accept lists and tagged commands for the user-supplied program to execute
- New commands added: LSCAN, LMAP, LXEQ, LDIFF, LINTR, LSUBS, LFILT, LWHL, LDWHL, LTWHL, LCPRD, LTAKE, LDROP, LCOMP, LDLTA, NL→I2
- Updated documentation for new and changed commands
 « Next Oldest | Next Newest »

User(s) browsing this thread: 1 Guest(s)