sub-.garray: Indexing for the garray class
In garray: Generalized Array Arithmetic for Ragged Arrays with Named Margins

Indexing along margins as usual [.array, and along subdim.

## S3 method for class 'garray'
...[drop=TRUE]
#`[.garray`(..., drop=TRUE)
#x[i]
#x[i,j,...,drop=TRUE]
#x[m]
#x[l]
#x[M=i,N=j,...]

#\method{[}{garray}(...) <- value
#`[<-.garray`(..., value)
#x[i] <- value
#x[i,j,...] <- value
#x[m] <- value
#x[l] <- value
#x[M=i,N=j,...] <- value

`drop`	Whether indeces where 1==dim are removed. Different from R's native `[`, a garray will become a garray or scalar, never a vector.
`value`	An array or a scalar.
`x`	A generalized array from which elements are extracted or replaced.
`i, j, m, l, M, N, ...`	In addition to the native styles (`i`, `j`, etc.) accepted by `[`, can be: a matrix `m` with column names, which (`colnames(m)`) is a permutation of margins of the array. an list `l <- list(i,j,...)`, can be unnamed or named, where NULL means to select all; arguments with names (`M`, `N`, etc), where `NULL` and missing means to select all. These extensions make indexing 3 times slower than native indexing. Since it is hard to assign MissingArg in list(), at the moment MissingArg is only safe in native R subsettting style. Using NULL to select all like MissingArg is actually not consistent in semantic of other uses of NULL. As shown by what `c()` returns, NULL is a generalized form of `logical(0)`, `integer(0)`, and `character(0)`, all of which means to select none when indexing. So take care of NULL if indexing with variables.

mm <- matrix(c(1:3,1), 2, 2, dimnames=list(NULL, c("B","A")))
a <- garray(1:27, c(A=3,B=9), sdim=list(AA=c(a=2,b=1),BB=c(a=3)))
b <- a[mm]
c1 <- a[B=1:2,A=NULL]
c2 <- a[B=1:2,A=]
c3 <- a[B=1:2]
c4 <- a[list(B=1:2)]
c5 <- a[list(B=1:2,A=NULL)]
c6 <- a[list(NULL,1:2)]
d1 <- a[,] ; d1[B=1:2,A=NULL]       <- c1*10
d2 <- a[,] ; d2[B=1:2,A=]           <- c1*10
d3 <- a[,] ; d3[B=1:2]              <- c1*10
d4 <- a[,] ; d4[list(B=1:2)]        <- c1*10
d5 <- a[,] ; d5[list(B=1:2,A=NULL)] <- c1*10
d6 <- a[,] ; d6[B=1:2,A=NULL] <- 1
d7 <- a[,] ; d7[mm] <- 1000
d8 <- a[,] ; d8[mm] <- 1:2*1000
e1 <- a[AA=1,drop=FALSE]
e11 <- a[AA=c(1,1),drop=FALSE]
e2 <- a[AA="b",drop=FALSE]
ebb <- a[AA=c("b","b"),drop=FALSE]
e3 <- a[,] ; e3[AA="b"] <- e2*10
e33 <- a[,] ; e33[AA=c("b","b")] <- c(e2*0.1, e2*100)
# Work in the same manner of `e33[c(3,3),] <- c(e2*0.1, e2*100)`.
e4 <- a[A=c(TRUE,FALSE,FALSE),drop=FALSE]
e5 <- a[A=TRUE,drop=FALSE]
e6 <- a[B=c(TRUE,FALSE,FALSE),drop=FALSE]
e7 <- a[AA=TRUE,drop=FALSE]
e8 <- a[AA=c(TRUE,FALSE),drop=FALSE]