--- title: "Using beachmat to read data from R matrices in C++" author: "Aaron Lun" package: beachmat output: BiocStyle::html_document vignette: > %\VignetteIndexEntry{Using beachmat to read data from R matrices in C++} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} --- ```{r, echo=FALSE, results="hide", message=FALSE} require(knitr) opts_chunk$set(error=FALSE, message=FALSE, warning=FALSE) ``` # Overview of the input API This document describes the use of the `r Biocpkg("beachmat")` API for accessing data in R matrices. We will demonstrate the API on numeric matrices, though same semantics are used for matrices of other types (e.g., logical, integer, character). First, we include the relevant header file: ```cpp #include "beachmat/numeric_matrix.h" ``` A double-precision matrix object `dmat` is handled in C++ by passing the `SEXP` struct from `.Call` to `create_numeric_matrix`: ```cpp // returns a std::unique_ptr object auto dptr = beachmat::create_numeric_matrix(dmat); ``` This creates a unique pointer that points to an object of the `numeric_matrix` base class. The exact derived class that is actually instantiated depends on the type of matrix in `dmat`, though the behaviour of the user-level functions are not affected by this detail. **Additional notes:** - The `auto` keyword just avoids the need to write the full type of the returned pointer, which is `std::unique_ptr`. We use unique pointers to control ownership and smoothly handle destruction and memory deallocation at the end of the function. - The API will happily throw exceptions of the `std::exception` class, containing an informative error message. These should be caught and handled gracefully by the end-user code, otherwise a segmentation fault will probably occur. See the error-handling mechanism in `r CRANpkg("Rcpp")` for how to deal with these exceptions. # Querying matrix information The `get_nrow()` method returns the number of rows (referred to below as `nrow`) in the matrix as a `size_t`. ```cpp dptr->get_nrow(); ``` The `get_ncol()` method returns the number of columns (referred to below as `ncol`) in the matrix as a `size_t`. ```cpp dptr->get_ncol(); ``` The `get_matrix_type()` method returns a `matrix_type` value specifying the specific matrix representation that is pointed to by `dptr`. This is an enumeration type that can be tested against constants like `beachmat::SIMPLE` or `beachmat::SPARSE`. ``` dptr->get_matrix_type(); ``` The `yield()` method returns a `Rcpp::RObject` containing the original R matrix that was used to create `dptr`. ``` dptr->yield() ```` # Extracting data ## From columns The `get_col()` method takes an iterator `in` to a _Rcpp_ vector and fills it with values at column `c`. There should be at least `nrow` accessible elements, i.e., `*in` and `*(in+nrow-1)` should be valid entries. ```cpp dptr->get_col(/* size_t */ c, /* Rcpp::Vector::iterator */ in); ``` Extraction of a range of the column can be specified with the `first` and `last` arguments. This will fill `in` with values at column `c` from row `first` to `last-1`. There should be at least `last-first` accessible elements, i.e., `*in` and `*(in+last-first-1)` should be valid entries. ```cpp dptr->get_col(/* size_t */ c, /* Rcpp::Vector::iterator */ in, /* size_t */ first, /* size_t */ last); ``` No value is returned by either of these methods. Note that `c` should be a zero-indexed integer in `[0, ncol)`. Similarly, both `first` and `last` should be in `[0, nrow]` and zero-indexed, with the additional requirement that `last >= first`. ## From rows The `get_row()` method takes an iterator `in` to a _Rcpp_ vector and fills it with values at row `r`. There should be at least `ncol` accessible elements, i.e., `*in` and `*(in+ncol-1)` should be valid entries. ```cpp dptr->get_row(/* size_t */ r, /* Rcpp::Vector::iterator */ in); ``` Extraction of a range of the row can be specified with the `first` and `last` arguments. This will fill `in` with values at row `r` from column `first` to `last-1`. There should be at least `last-first` accessible elements, i.e., `*in` and `*(in+last-first-1)` should be valid entries. ```cpp dptr->get_row(/* size_t */ r, /* Rcpp::Vector::iterator */ in, /* size_t */ first, /* size_t */ last); ``` No value is returned by either of these methods. Again, `r` should be a zero-indexed integer in `[0, nrow)`. Both `first` and `last` should be in `[0, ncol]` and zero-indexed, with the additional requirement that `last >= first`. ## From individual cells The `get()` method returns a double-precision value at the matrix entry for row `r` and column `c`. Both `r` and `c` should be zero-indexed integers in `[0, nrow)` and `[0, ncol)` respectively. ```cpp dptr->get(/* size_t */ r, /* size_t */ c) ``` ## Type conversions If the object `in` is a `Rcpp::NumericVector::iterator` instance, matrix entries will be extracted as double-precision values. If it is a `Rcpp::IntegerVector::iterator` instance, matrix entries will be extracted as integers with implicit conversion. It is also _possible_ to use a `Rcpp::LogicalVector::iterator`, though this will not behave as expected. Double-to-integer conversion is performed such that values in `(-1, 1)` are converted to integer `0`. This would be interpreted as a logical `FALSE`, which is incorrect for non-zero double-precision values. ## Fast constant access ### From dense matrices The `get_const_col()` method returns an iterator to a _Rcpp_ vector pointing to `first` row of column `c`. The arguments are the same as `get_col()` with `work` being equivalent to `in`. (Note that `first` and `last` do not have to be specified and will default to the entire column.) The type of `work` and the output iterator must correspond to the data type of the matrix - in this case, both should be `Rcpp::NumericVector::iterator` objects. ```cpp dptr->get_const_col(/* size_t */ c, /* Rcpp::NumericVector::iterator */ work, /* size_t */ first, /* size_t */ last); ``` For simple/dense matrices, this method is more efficient than `get_col()` as it returns the iterator without needing to copy data into `work`. For other matrices, this function simply calls `get_col()` to copy the data into the vector starting at `work`, and then returns an iterator equal to `work`. Thus, for general use, `work` must point to a writeable block of memory, even if it does not get used when `dptr` points to a simple/dense matrix. **Additional notes:** - Calling processes should consider the vector starting at `work` to be read-only after a call to `get_const_col()`. In particular, the underlying data will be modified if `get_const_col()` is called again with the same `work`. This has some consequences if a process calls `get_const_col()` multiple times to obtain more than one iterators for further computation. In such cases, developers should ensure that each `get_const_col()` call uses a `work` pointing to a different vector. ### From sparse matrices The `get_const_col_indexed()` method returns a `const_col_indexed_info` object, which is a tuple consisting of: 1. A `size_t`, specifying the number of entries in column `c` from rows `[first, last)`. 2. A `Rcpp::IntegerVector::iterator`, pointing to a vector containing the row index for each entry in column `c` from rows `[first, last)`. 3. A `Rcpp::NumericVector::iterator`, pointing to a vector containing the value of each entry in column `c` from rows `[first, last)`. The data type of `work` and the output iterator in (3) must correspond to that of the matrix - in this case, both should be `Rcpp::NumericVector::iterator` objects. Again, `first` and `last` do not have to be specified and will default to the entire column. ```cpp dptr->get_const_col_indexed(/* size_t */ c, /* Rcpp::NumericVector::iterator */ work, /* size_t */ first, /* size_t */ last); ``` This method is quite efficient for `dgCMatrix` objects, as it will directly return iterators to the indices and values of the _non-zero entries only_. No copying is performed, and the zero values do not have to be explicitly generated as they would be in `get_col`. For all other matrices, `get_const_col` is called^[The comments above regarding the read-only nature of `work` also apply here.] and iterator (2) is pointed at an internal array of consecutive integers (which should be treated as read-only). # Generalizing to other matrices ## Other data types To create logical, integer and character matrices, include the following header files: ```cpp #include "beachmat/logical_matrix.h" #include "beachmat/integer_matrix.h" #include "beachmat/character_matrix.h" ``` The dispatch function changes correspondingly for logical matrix `lmat`, integer matrix `imat` or character matrix `cmat`. Each function creates a unique pointer to a `*_matrix` of the appropriate type. ```cpp // creates a std::unique_ptr auto lptr=beachmat::create_logical_matrix(lmat); // creates a std::unique_ptr auto iptr=beachmat::create_integer_matrix(imat); // creates a std::unique_ptr auto cptr=beachmat::create_character_matrix(cmat); ``` Equivalent methods are available for each matrix type with appropriate changes in type. For integer and logical matrices, `get` will return an integer. `in` can be any type previously described for `numeric_matrix` objects^[Note the caveats previously discussed for `Rcpp::LogicalVector` iterators.]. `work` should be an `iterator` to a `Rcpp::IntegerVector` for integer matrices or a `Rcpp::LogicalVector` for logical matrices. Similar changes apply to the iterator in the tuple from `get_const_col_indexed()`. For character matrices, all iterators should be of type `Rcpp::StringVector::iterator`, and `get` will return a `Rcpp::String`. **Additional notes:** - If `in` is a `Rcpp::LogicalVector::iterator` for `integer_matrix` instances, integer values are not coerced to `{0, 1}` when they are assigned to `*in`. Thus, even though the interpretation is correct, the vector produced will not be equivalent to the result of an `as.logical` call. As a general rule, it is unwise to use `Rcpp::LogicalVector::iterator`s for anything other than `logical_matrix` access. - When accessing `character_matrix` data, we do not return raw `const char*` pointers to the C-style string. Rather, the `Rcpp::String` class is used as it provides a convenient wrapper around the underlying `CHARSXP`. This ensures that the string is stored in R's global cache and is suitably protected against garbage collection. ## Alternative matrix representations The following matrix classes are supported: - numeric: `matrix`, `dgeMatrix`, `dgCMatrix`, `dspMatrix`, `RleMatrix`, `HDF5Matrix`, `DelayedMatrix` - integer: `matrix`, `RleMatrix`, `HDF5Matrix`, `DelayedMatrix` - logical: `matrix`, `lgeMatrix`, `lgCMatrix`, `lspMatrix`, `RleMatrix`, `HDF5Matrix`, `DelayedMatrix` - character: `matrix`, `RleMatrix`, `HDF5Matrix`, `DelayedMatrix` Additional classes may be added on request. As a general rule, if a matrix-like object can be stored in a `SummarizedExperiment` class (from the `r Biocpkg("SummarizedExperiment")` package), the API should be able to handle it. Please contact the maintainers if you have a class that you would like to see supported. **Additional notes:** - `DelayedMatrix` objects can be directly parsed by the API if the only delayed operations involve transposition or subsetting. Otherwise, they are realized in a chunk-by-chunk manner, using methods in the `r Biocpkg("DelayedArray")` package. The granularity of the chunking is determined using the `defaultGrid` function. - For numeric matrices, _beachmat_ does not support higher-level matrix operations such as addition, multiplication or various factorizations. Rather, the `yield` method can be used to obtain the original `Rcpp::RObject` for input to `r CRANpkg("RcppArmadillo")` or `r CRANpkg("RcppEigen")`. This functionality is generally limited to base matrices, though there is also limited support for sparse matrices in these libraries. # Cloning for parallelization The `clone()` method returns a unique pointer to a `numeric_matrix` instance of the same type as that pointed to by `dptr`. ``` dptr->clone(); ``` This is useful as direct use of the `r Biocpkg("beachmat")` API is not thread-safe for simultaneous calls to the `get` methods from different threads. Some methods use cached class members for greater efficiency, and simultaneous calls will cause race conditions. It is the responsibility of the calling function to coordinate data access across threads. To this end, the `clone` method can be called to generate a unique pointer to a _new_ `*_matrix` instance, which can be used concurrently in another thread. This is fairly cheap as the underlying matrix data are not copied. An example of parallelized `r Biocpkg("beachmat")` code using OpenMP might look like this: ```cpp #pragma omp parallel { beachmat::numeric_matrix* rptr=NULL; std::unique_ptr uptr=nullptr; if (omp_get_thread_num()==0) { rptr=dptr.get(); } else { uptr=dptr->clone(); rptr=uptr.get(); } size_t col; const size_t NC=rptr->get_ncol(); Rcpp::NumericVector output(rptr->get_nrow()); #pragma omp for schedule(static) for (col=0; colget_col(col, output.begin()); } } ``` The start of the parallel region uses the existing `dptr` in the master thread and clones a new matrix in the other threads. The parallelized `for` loop then uses `rptr` to avoid race conditions in cached variables. Note that a static schedule may be faster than other schedule types, as several of the matrix implementations in `r Biocpkg("beachmat")` are optimized for consecutive row/column access.