robinhankin / frab

Hello everyone, my name is Robin and I work on computational statistics, mostly in R

frab's Issues

replacement method bug

library("frab")
#> 
#> Attaching package: 'frab'
#> The following object is masked from 'package:graphics':
#> 
#>     symbols
(x <- rfrab())
#> A frab object with hash a391ff3181befefb96022c7098781d567820b3c0 and entries
#> d e g h 
#> 8 7 3 4
x[] <- 3
#> Error in .local(x, i, j, ..., value): replacement operation not defined in this case

^{Created on 2023-07-11 with reprex v2.0.2}

it would seem that x[] <- 3 has a clear expectation.

bug in extraction

#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
x <- frab(c(x=4,y=6,z=8))
x[is.na(x)]
#> The empty frab object
x[!is.na(x)]
#> The empty frab object

last line above is clearly a bug, the result should be x.

bug in addition

suppressWarnings(library("frab"))
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
x <- frab(c(a=1,b=-4,c=2,d=5,e=6,f=-9))
x[x>5] + 1000
#> Error in as.frab(e2): argument not recognised; did you mean to pass a _named_ numeric vector?

Above I would expect the sum to be defined but it gives an inappropriate error

Function `frab()` really should be able to take a named _logical_ vector

and it can't:

library("frab")
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
frab(c(a=TRUE,b=TRUE,c=FALSE,d=TRUE))
#> Error in frab(c(a = TRUE, b = TRUE, c = FALSE, d = TRUE)): is.namedvector(x) is not TRUE

^{Created on 2023-07-26 with reprex v2.0.2}

pmin is not used

In src/sparsetable_ops.cpp, function List sparsetable_pmin() is not used: in R, pmin is defined in terms of pmax(),
specifically we have

pmin_pair_sparsetable <- function(F1,F2){-pmax_pair_sparsetable(-F1,-F2)}`

bug in `sparsetable()`

suppressMessages(library("frab"))
(x <- sparsetable(matrix(letters[1:24],ncol=3),1))
#>            val
#>  a i q  =    1
#>  b j r  =    0
#>  c k s  =    0
#>  d l t  =    0
#>  e m u  =    0
#>  f n v  =    0
#>  h p x  =    3
dput(x)
#> new("sparsetable", index = structure(c("a", "b", "c", "d", "e", 
#> "f", "h", "i", "j", "k", "l", "m", "n", "p", "q", "r", "s", "t", 
#> "u", "v", "x"), dim = c(7L, 3L)), values = c(1, 4.65015486152201e-314, 
#> 4.65625948802518e-310, 4.65626113840799e-310, 4.65626113840245e-310, 
#> 4.94065645841247e-324, 3))

^{Created on 2023-08-09 with reprex v2.0.2}

Above the values show crazy small values because the proper vector was not sent to sparsetable()

inefficiency in equality check

Look at the check for equality in frab.cpp:

bool equal2_samesize(frab F1, frab F2){
  for (auto it=F1.begin(); it != F1.end(); ++it){
      const string symbol = it->first;
      if(F1[symbol] != F2[symbol]){  // meat 1
	return false;
      } else {
	F2.erase(symbol);
      }
  }
  return F2.empty();
}

This is inefficient, given that F1 and F2 are the same size. If they are the same size, we can just iterate through F1 and return false if it finds a difference and true if the loop completes, with no need for an erase operation. I think the erase step was originally there to guard against incorrectly returning true if F1 was a strict subset of F2 (and every entry of F1 agreed with the corresponding entry of F2).

replacement with NA bug

library("frab")
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
set.seed(0)
x <- rfrab()
x
#> A frab object with entries
#> a b c d g i 
#> 3 6 1 5 7 5
x["c"] <- 100
x
#> A frab object with entries
#>   a   b   c   d   g   i 
#>   3   6 100   5   7   5
x["c"] <- NA
#> Error in .local(x, i, j, ..., value): replacement operation not defined in this case

Above x["c"] <- NA has a well-defined meaning and I would expect it to work.

bug in `pmax()`

library("frab")
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
pmax(rfrab(),rfrab(),rfrab())
#> A frab object with entries
#>  a  b  c  d  e  f  g  h  i 
#> 13  4  4  1  4  3  8  3  7
pmax(rfrab(),rfrab(),rfrab(),rfrab())
#> Error in pmax_pair(...): unused argument (new("frab", x = c(7, 4, 1, 4, 4, 1, 6)))

^{Created on 2023-07-26 with reprex v2.0.2}

colnames of index matrix get lost

suppressMessages(library("frab"))
I <- matrix(sample(letters[1:4],15,replace=T),ncol=3)
colnames(I) <- state.abb[1:3]
I
#>      AL  AK  AZ 
#> [1,] "b" "b" "c"
#> [2,] "d" "a" "b"
#> [3,] "c" "b" "a"
#> [4,] "c" "a" "d"
#> [5,] "b" "a" "b"
sparsetable(I)
#>            val
#>  b a b  =    1
#>  b b c  =    1
#>  c a d  =    1
#>  c b a  =    1
#>  d a b  =    1

^{Created on 2023-08-11 with reprex v2.0.2}

sparsetable comparison methods not well documented

Just looking at issue #25 it is not clear why a ==3 returns a disord object and not some kind of Boolean sparsetable.
At least, the Rd files don't make this clear.

from CRAN

Thanks,

Please do not start the description with "This package", package name,
title or similar.

Please add \value to .Rd files regarding exported methods and explain
the functions results in the documentation. Please write about the
structure of the output (class) and also what the output means. (If a
function does not return a value, please document that too, e.g.
\value{No return value, called for side effects} or similar)
Missing Rd-tags:
print.Rd: \value
rfrab.Rd: \value
zero.Rd: \value

Please fix and resubmit.

Best,
Benjamin Altmann

missing replacement methods

It seems to be quite difficult to replace values of a frab object. Given a frab with some negative values, it is quite easy to replace
the negative values with a particular scalar:

library("frab")
set.seed(0)
(x <- rfrab(10,-2:2))
#> A frab object with entries
#>  c  d  e  g  i 
#> -2 -1  2  1  2
x[x<0] <- 0
x
#> A frab object with entries
#> e g i 
#> 2 1 2

However, something like abs(x) seems to be much harder:

library("frab")
set.seed(0)
(x <- rfrab(10,-2:2))
#> A frab object with entries
#>  c  d  e  g  i 
#> -2 -1  2  1  2
abs(x)
#> Error in abs(x): non-numeric argument to mathematical function
x[x<0]
#> A frab object with entries
#>  c  d 
#> -2 -1
x[x<0] <- -x[x<0]
#> Error in .local(x, i, j, ..., value): replacement operation not defined in this case

Above, x[x<0] <- -x[x<0] would seem to have a very reasonable expectation, but it does not work, due to a lack of

setReplaceMethod("[",signature(x="frab",i="disord",j="missing",value="frab"),...)

which would seem to be relatively straightforward to implement. Actually the best way I have found in existing package idiom to achieve abs(x) is this:

suppressMessages(library("frab"))
set.seed(0)
(x <- rfrab(10,-2:2))
#> A frab object with entries
#>  c  d  e  g  i 
#> -2 -1  2  1  2
frab(abs(as.namedvector(x)))
#> A frab object with entries
#> c d e g i 
#> 2 1 2 1 2

Part of the problem is that we do not have a values<-() method. But I need to make values() and values<-() in the disordR package generic first.

not operator for `is.na()`

The "not" operator does not work for is.na():

suppressMessages(library("frab"))
(x <- frab(c(x=5,y=1,z=2)))
#> A frab object with entries
#> x y z 
#> 5 1 2
x["y"] <- NA
is.na(x)
#> A disind object with hash eec061606d73bf83eced3b5d7d31b5b14459f96f and 1 (implementation-specific) elements
!is.na(x)
#> Error in !is.na(x): invalid argument type

Now I don't think that "!is.na(x)" should work, but it is not documented clearly.

rfrabbb() infelicity

rfrabbb(n=10,i=5) takes a very long time to run. This is because of the expand.grid() constuction. Something like

replicate(5,paste(sample(letters,5,replace=TRUE),collapse=""))
# [1] "jydlp" "kumyb" "xwxmn" "tqvsd" "twiwk"

might be better.

sparsetable inequality (!=) not implemented

library("frab")
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
(x <- rspar2())
#>    bar
#> foo A  B  C D  E  F
#>   a 4 14  0 0  0  7
#>   b 0  5 13 0  0  9
#>   c 6  1  0 0  0  3
#>   d 0  0  0 2  0 10
#>   e 0  8 12 0  0  0
#>   f 0 15  0 0 11  0
x == 3
#> A disord object with hash 5a45a89691589101cbf771e946c210366eeda5fd and elements
#>  [1] FALSE FALSE FALSE FALSE FALSE FALSE FALSE FALSE  TRUE FALSE FALSE FALSE
#> [13] FALSE FALSE FALSE
#> (in some order)
x != 3
#> Error in x != 3: Comparison operator "!=" not implemented in this case

frab == named vector

suppressWarnings(library("frab"))
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
frab(c(a=1,b=2,c=3)) == c(a=1,b=2,c=3)
#> 
#> disordR discipline error in:
#> values(e1) == e2
#> Error in check_matching_hash(e1, e2, match.call()): 
#> cannot combine disord object with hash code 471a46a1167406f770fe039498ad8c1da1f59f6d with a vector

Above, the == operator has a clear expectation (TRUE) and it would be reasonable (?) to coerce the RHS to a frab before attempting the coercion

replacement methods for `names`

Wouldn't it be nice to explicitly reset the names of a frab object:

library("frab")
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
(a <- frab(c(my_flob = 3, my_banana=1, my_float = 22)))
#> A frab object with entries
#> my_banana  my_float   my_flob 
#>         1        22         3
gsub("my_","",names(a))
#> A disord object with hash c74a27de91c41202410d366d87f5777a00c3d172 and elements
#> [1] "banana" "float"  "flob"  
#> (in some order)
names(a) <- gsub("my_","",names(a))
#> Error in names(a) <- gsub("my_", "", names(a)): invalid to use names()<- on an S4 object of class 'frab'

(a_desired <- frab(c(flob = 3, banana=1, float = 22)))
#> A frab object with entries
#> banana  float   flob 
#>      1     22      3

Above we see that names<-() is not implemented. It would be nice to have replacement methods defined so that
names(a) <- gsub("my_","",names(a)) gives a_desired.

In passing, I would say I am delighted and surprised that gsub() takes, and returns, a disord object. This should be documented somewhere.

x[is.na(x)] fails if empty

suppressMessages(library("frab"))
(x <- frab(c(x=5,y=1,z=2)))
#> A frab object with entries
#> x y z 
#> 5 1 2
x["y"] <- NA
x[is.na(x)]
#> A frab object with entries
#>  y 
#> NA
x[is.na(x)] <- 33
x[is.na(x)]
#> Error in .local(x, i, j = j, ..., drop): not implemented

Above, the first call to x[is.na(x)] works fine, extracting element y. But the second fails.

pmin() and pmax()

library("frab")
library("disordR")
#> Loading required package: Matrix
#> 
#> Attaching package: 'disordR'
#> The following object is masked from 'package:frab':
#> 
#>     values
(x <- frab(c(a=1,b=2,c=3)))
#> A frab object with entries
#> a b c 
#> 1 2 3
(y <- frab(c(b=4,c=1,d=144)))
#> A frab object with entries
#>   b   c   d 
#>   4   1 144
pmin(x,y)
#> Warning in is.na(mmm): is.na() applied to non-(list or vector) of type 'S4'
#> Warning in is.na(each): is.na() applied to non-(list or vector) of type 'S4'
#> Error in mmm > each: comparison '">"' not for frabs

Above, I would expect pmin(x,y) to give frab(c(a=1,b=2,c=1)) [NB d has disappeared]

bug in replacement

library("frab")
library("disordR")
#> Loading required package: Matrix
#> 
#> Attaching package: 'disordR'
#> The following object is masked from 'package:frab':
#> 
#>     values
x <- frab(c(a=1,b=2,c=3,i=4))
x
#> A frab object with entries
#> a b c i 
#> 1 2 3 4
x[] <- c(4,6,77,6)
x
#> A frab object with entries
#>  a  b  c  i 
#>  4  6 77  6

above I would expect x[] <- c(4,6,77,6) to return a disord discipline error

powers and symbols terminology is outdated

I've realised that the main purpose of the frab package is as an alternative interpretation of named vectors.
Although powers and symbols are suitable if one considers the Free Abelian Group as a special case of the Free Group [as implemented by freegroup], they are somewhat confusing and weird if one is looking for a method to manipulate named vectors.
we should have values and names instead.

summary() method unhelpful

library("frab")
#> 
#> Attaching package: 'frab'
#> The following objects are masked from 'package:base':
#> 
#>     pmax, pmin
x <- rfrab()
x
#> A frab object with entries
#> b c d e f g h i 
#> 6 2 3 3 4 5 3 5
summary(x)
#> Length  Class   Mode 
#>      8   frab     S4

^{Created on 2023-08-02 with reprex v2.0.2}

infelicity in comparison methods

library("frab")
#> 
#> Attaching package: 'frab'
#> The following object is masked from 'package:graphics':
#> 
#>     symbols
(x <- frab(c(x=5,a=3,d=6)))
#> A frab object with hash 25076b3f8c2dbd97d18146872eb27d08034b8dab and entries
#> a d x 
#> 3 6 5
x ==  3
#> A disord object with hash 25076b3f8c2dbd97d18146872eb27d08034b8dab and elements
#> [1] "a"
#> (in some order)

^{Created on 2023-07-11 with reprex v2.0.2}

disordR discipline violation

suppressPackageStartupMessages(library("frab"))
x <- frab(c(a=-3,b=-1,c=-2,d=1,f=-1,h=1,i=-1))
x
#> A frab object with entries
#>  a  b  c  d  f  h  i 
#> -3 -1 -2  1 -1  1 -1

x[x>0]
#> A frab object with entries
#> d h 
#> 1 1

x[x>0] + 1:2
#> A frab object with entries
#> d h 
#> 2 3

Above we see a bug, surely x[x>0] + 1:2 should raise a disordR error?

bug in replacement methods

library("frab")
x <- frab(c(a=1,b=2,c=3,i=4))
x
#> A frab object with entries
#> a b c i 
#> 1 2 3 4
x[x>2] <- 33
x
#> A frab object with entries
#>  1  2  a  b  c  i 
#> 33 33  1  2 33 33

^{Created on 2023-07-13 with reprex v2.0.2}

null sparstable object fail

suppressMessages(library("frab"))
a <- rspar()
a-a
#> Error in eval(expr, envir, enclos): Not compatible with STRSXP: [type=NULL].

^{Created on 2023-08-12 with reprex v2.0.2}