Note to self: the specification of the output distance vector in the vignette is wrong

Here's a that will compile the C++ code in the vignette: <div class="highli

You could of course just write it as a vignette too :) <div class="snippet-clipboa

Fix vignette description of output distance vector about rcppannoy HOT 6 CLOSED

eddelbuettel commented on May 31, 2024

Fix vignette description of output distance vector

from rcppannoy.

Comments (6)

LTLA commented on May 31, 2024 1

In fact, I wonder whether it would be worth putting this into the package as an official unit test. This would check that the vignette's advice is correct across versions and across all CRAN systems.

from rcppannoy.

eddelbuettel commented on May 31, 2024

Hm, I just tried knitr::purl() and it failed returning an essentially empty file.

Looking more closely I see that your code chunks are just pandoc chunks for hightlighting (a mode I use myself all the time too). Short of cooking up a simple regexp expression, you can probably ask pandoc to extract these. But you'd have to look at how.

from rcppannoy.

LTLA commented on May 31, 2024

Here's a script that will compile the C++ code in the vignette:

# Identify all 'cpp' code chunks.
X <- readLines("rcppannoy.Rmd")
starts <- grep("^```cpp$", X)
ends <- grep("^```$", X)

I <- findInterval(ends, starts)
keep <- !duplicated(I) & I>0L

code.I <- mapply(seq, from=starts+1L, to=ends[keep]-1L)
code.chunks <- X[unlist(code.I)]

# Add preprocessor statements, variables necessary for compilation.
end.pp <- grep("MyAnnoyIndex;$", code.chunks)
total <- c(
    head(code.chunks, end.pp),
    "#include <vector>",
    "#include <algorithm>",
    "// [[Rcpp::depends(RcppAnnoy)]]",
    "// [[Rcpp::export]]", 
    "Rcpp::IntegerVector testfun(Rcpp::NumericMatrix mat) {",
    "int c=0, K=5;",
    "std::vector<ANNOYTYPE> stub(mat.ncol());",
    "ANNOYTYPE* query=stub.data();",
    tail(code.chunks, -end.pp),
    "return Rcpp::IntegerVector(neighbor_index.begin(), neighbor_index.end());",
    "}"
)   

# Compile and execute as a test.
total <- paste(total, collapse="\n")
library(Rcpp)
out <- sourceCpp(code=total)
testfun(matrix(0, 10, 5))