R/3.0_read_omics.R

In bhagwataditya/importomics: Unified statistal Modeling of Omics Data

#==============================================================================
#
#                       is_excel_file
#                       is_fixed_col_file
#                       nrows
#                       ncols
#
#==============================================================================

is_excel_file <- function(file){
    stri_detect_fixed(
            file_ext(file), 'xls', case_insensitive = TRUE)
}

nrows <- function(x, sheet=1){
    if (is_excel_file(x)){
        nrow(read_excel(
                x, sheet = sheet, .name_repair = 'minimal', col_names = FALSE))
    } else {
        length(readLines(x, warn=FALSE))
    }
}


ncols <- function(x, sheet=1){
    if (is_excel_file(x)){
        ncol(read_excel(
            x, sheet=sheet, .name_repair = 'minimal', col_names = FALSE))
    } else {
        max(count.fields(x, quote = "", sep = '\t'))
    }
}

#' Is this a fixed column file?
#' @param file string: file name
#' @return TRUE or FALSE
#' @examples
#' # SOMASCAN
#'     file <- system.file('extdata/atkin.somascan.adat', package = 'autonomics')
#'     is_fixed_col_file(file)
#'
#' # METABOLON
#'     file <- system.file('extdata/atkin.metabolon.xlsx', package = 'autonomics')
#'     is_fixed_col_file(file)
#' @noRd
is_fixed_col_file <- function(file){
    if (is_excel_file(file)){ TRUE
    } else {
        fields <- count.fields(file, quote = "", sep = '\t')
        all(fields==fields[1])
    }
}

#==============================================================================
#
#                           extract_rectangle
#
#==============================================================================

#' Extract rectangle from omics file, data.table, or matrix
#'
#' @param x          omics datafile or datatable
#' @param rows       numeric vector
#' @param cols       numeric vector
#' @param verbose    logical
#' @param transpose  logical
#' @param drop       logical
#' @param sheet      numeric or string
#' @param ...        allow for S3 method dispatch
#' @return matrix
#' @examples
#' # FROM FILE: extract_rectangle.character
#' #=======================================
#'    x <- system.file('extdata/atkin.metabolon.xlsx', package = 'autonomics')
#'    extract_rectangle(x, rows = 11:30, cols = 15:81, sheet = 2)[ 1:3, 1:3 ]  # exprs
#'    extract_rectangle(x, rows = 11:30, cols = 2,     sheet = 2)[ 1:3,     ]  # fids
#'    extract_rectangle(x, rows = 4,     cols = 15:81, sheet = 2)[    , 1:3 ]  # sids
#'    extract_rectangle(x, rows = 10:30, cols = 1:14,  sheet = 2)[ 1:3, 1:3 ]  # fdt
#'    extract_rectangle(x, rows = 1:10,  cols = 14:81, sheet = 2, transpose = TRUE)[1:3, 1:3] # sdt
#'
#' # FROM MATRIX: extract_rectangle.matrix
#' #======================================
#'    x <- system.file('extdata/atkin.metabolon.xlsx', package = 'autonomics') %>% extract_rectangle(sheet = 2)
#'    extract_rectangle(x, rows = 11:30,  cols = 15:81, sheet = 2)[ 1:3, 1:3 ]  # exprs
#'    extract_rectangle(x, rows = 11:30,  cols = 2,     sheet = 2)[ 1:3,     ]  # fids
#'    extract_rectangle(x, rows = 4,      cols = 15:81, sheet = 2)[    , 1:3 ]  # sids
#'    extract_rectangle(x, rows = 10:30,  cols = 1:14,  sheet = 2)[ 1:3, 1:3 ]  # fdt
#'    extract_rectangle(x, rows = 1:10,   cols = 14:81, sheet = 2, transpose = TRUE)[1:3, 1:3] # sdt
#' @export
extract_rectangle <- function(x, ...){
    UseMethod('extract_rectangle')
}

#' @rdname extract_rectangle
#' @export
extract_rectangle.character <- function(x, rows=seq_len(nrows(x, sheet=sheet)),
    cols = seq_len(ncols(x, sheet=sheet)), verbose = FALSE, transpose = FALSE,
    drop = FALSE, sheet = 1, ...
){
    # Assert
    assert_all_are_existing_files(x)
    assert_is_numeric(rows)
    assert_is_numeric(cols)
    row1 <- min(rows); rown <- max(rows)
    col1 <- min(cols); coln <- max(cols)

    # Read file
    if (verbose) message('\t\tRead ',
                if (sheet==1) '' else paste0("sheet '", sheet, "' of"), file)
    dt <- if (is_excel_file(x)){
            data.table(read_excel(
                x,
                sheet       = sheet,
                col_names   = FALSE,
                range       = sprintf('R%dC%d:R%dC%d', row1, col1, rown, coln),
                .name_repair = 'minimal'))
        } else {
            fread(
                x,
                na.strings = "",
                header     = FALSE,
                integer64  = 'numeric',
                skip       = row1-1,
                nrows      = 1+rown-row1,
                select     = seq(col1,coln)) }

    # Extract rectangle
    extract_rectangle.data.table(dt, transpose = transpose, drop = drop)

}

# Extract row
extract_dt_row <- function(dt, i) unname(as.matrix(dt[i,])[1,])
extract_dt_col <- function(dt, i) dt[[i]]


#' @rdname extract_rectangle
#' @export
extract_rectangle.data.table <- function(
    x,
    rows = seq_len(nrow(x)),
    cols = seq_len(ncol(x)),
    transpose = FALSE,
    drop = FALSE,
    ...
){
    extract(x, rows, cols, with = FALSE) %>%
    as.matrix() %>%
    extract_rectangle.matrix(transpose = transpose, drop = drop)
}


#' @rdname extract_rectangle
#' @export
extract_rectangle.matrix <- function(
    x,
    rows      = seq_len(nrow(x)),
    cols      = seq_len(ncol(x)),
    transpose = FALSE,
    drop      = FALSE,
    ...
){
    rectangle <- x[rows, cols, drop = FALSE]
    if (transpose) rectangle %<>% t()
    if (drop) if (nrow(rectangle)==1 | ncol(rectangle)==1) rectangle %<>%
                                                        as.vector('character')
    rectangle
}

#==============================================================================
#
#                         extract_fdata
#                         extract_sdata
#
#==============================================================================

# Leave rownames(fdata1) empty: fids1 may contain non-valid values
# This happens in MaxQuant files, which sometimes contain missing rows
# (probably after opening in excell)
extract_fdata <- function(
    x, sheet, fids, fvar_rows, fvar_cols, fdata_rows, fdata_cols, transpose
){
    fdata1 <- data.frame(feature_id = fids, stringsAsFactors = FALSE)#,
                        #row.names  = fids) doesn't work with NA fids
    fdata_available <- !is.null(fvar_rows) & !is.null(fvar_cols)
    if (fdata_available){
        fvars1  <-  extract_rectangle(x,
                                    rows      = fvar_rows,
                                    cols      = fvar_cols,
                                    transpose = transpose,
                                    drop      = TRUE,
                                    sheet     = sheet)
        fdata1 %<>% cbind( extract_rectangle( x,
                                              rows       = fdata_rows,
                                              cols       = fdata_cols,
                                              transpose  = transpose,
                                              drop       = FALSE,
                                              sheet      = sheet) %>%
                            set_colnames(fvars1) %>%
                            data.frame(stringsAsFactors = FALSE, check.names = FALSE))
    }
    fdata1
}


# Leave rownames(sdata1) empty: sids may contain non-valid values
# This happens in SOMA files, where CLIENT_IDENTIFIER is not unique for
# calibrator and buffer samples
extract_sdata <- function(
    x, sheet, sids, svar_rows, svar_cols, sdata_rows, sdata_cols, transpose
){
    sdata1 <- data.frame(sample_id = sids, stringsAsFactors = FALSE)
    sdata_available <- !is.null(svar_rows) & !is.null(svar_cols)
    if (sdata_available){
        svars1 <- extract_rectangle(x,
                                    rows      = svar_rows,
                                    cols      = svar_cols,
                                    transpose =  transpose,
                                    drop      = TRUE,
                                    sheet     = sheet)
        sdata1 %<>% cbind(
                extract_rectangle(  x,
                                    rows       = sdata_rows,
                                    cols       = sdata_cols,
                                    transpose  = !transpose,
                                    drop       = FALSE,
                                    sheet      = sheet) %>%
                set_colnames(svars1) %>%
                data.frame(stringsAsFactors = FALSE, check.names = FALSE))
        sdata1 %<>% numerify()
    }
    sdata1
}


.is_numeric <- function(x)   all(grepl('^[0-9.]+$', x) | is.na(x) | x=='NA')
numerify   <- function(df){
    for (i in names(df)){
        if (.is_numeric(df[[i]]))  df[[i]] %<>% as.numeric()
    }
    df
}


#==============================================================================
#
#                         read_rectangles
#
#==============================================================================

#' @rdname read_rectangles
#' @export
.read_rectangles <- function(file, sheet = 1, fid_rows, fid_cols, sid_rows, 
    sid_cols, expr_rows, expr_cols, fvar_rows  = NULL, fvar_cols = NULL, 
    svar_rows = NULL, svar_cols = NULL, fdata_rows = NULL,  fdata_cols = NULL,
    sdata_rows = NULL, sdata_cols = NULL, transpose = FALSE, verbose = TRUE
){
# Read (in one go if fixed col file)
    assert_all_are_existing_files(file)
    assert_is_a_bool(transpose)
    if (verbose) message("\tRead: ", file)
    is_fixed_col <- is_fixed_col_file(file)
    x  <-   if (is_fixed_col){ extract_rectangle.character(file, sheet=sheet)
            } else {           file }
# Extract components
    fids1  <- extract_rectangle(x, rows = fid_rows, cols = fid_cols,
                        transpose = transpose, drop = TRUE, sheet = sheet)
    sids1  <- extract_rectangle(x, rows = sid_rows, cols = sid_cols,
                        transpose = transpose, drop = TRUE,  sheet = sheet)
    exprs1 <- extract_rectangle(x, rows = expr_rows, cols = expr_cols,
                        transpose = transpose, drop = FALSE, sheet = sheet)
    suppressWarnings(class(exprs1) <- 'numeric') # prevent NA warning
    fdata1 <- extract_fdata( x,   
                          sheet = sheet,
                           fids = fids1,
                     fvar_rows  = fvar_rows,
                      fvar_cols = fvar_cols,
                     fdata_rows = fdata_rows,
                     fdata_cols = fdata_cols,
                      transpose = transpose )
    sdata1 <- extract_sdata(x,   sheet = sheet, sids = sids1,
        svar_rows  = svar_rows,  svar_cols  = svar_cols,
        sdata_rows = sdata_rows, sdata_cols = sdata_cols, transpose = transpose)
# Rm features/samples with missing ids (eg MaxQuant: empty interspersed lines)
    idx <- !is.na(fids1)
    if (any(!idx)){
        if (verbose) message(
            "\t\tRm ", sum(!idx), " features with missing 'feature_id' values")
        fids1 <- fids1[idx]; fdata1 <- fdata1[idx,]; exprs1 <- exprs1[idx, ] }
    idx <- !is.na(sids1)
    if (any(!idx)){
        if (verbose) message(
            "\t\tRm ", sum(!idx), " samples with missing 'sample_id' values")
        sids1 <- sids1[idx]; sdata1 <- sdata1[idx, ]; exprs1 <- exprs1[, idx] }
# Name features/samples
    fids1 %<>% make.unique(); sids1 %<>% make.unique();
    rownames(exprs1) <- rownames(fdata1) <- fids1
    colnames(exprs1) <- rownames(sdata1) <- sids1
# Wrap into Sumexp and return
    object <- SummarizedExperiment(assays = list(exprs = exprs1))
    rowData(object) <- as(fdata1, 'DataFrame')
    colData(object) <- as(sdata1, 'DataFrame')
    metadata(object)$analysis  <-list(
        nfeatures = c(all = nrow(exprs1)), nsamples  = c(all = ncol(exprs1)))
    metadata(object)$file <- file
    object
}


#' Read omics data from rectangular file
#' @param file       string: name of text (txt, csv, tsv, adat) or
#'                           excel (xls, xlsx) file
#' @param sheet      integer/string: only relevant for excel files
#' @param fid_rows   numeric vector: featureid rows
#' @param fid_cols   numeric vector: featureid cols
#' @param sid_rows   numeric vector: sampleid rows
#' @param sid_cols   numeric vector: sampleid cols
#' @param expr_rows  numeric vector: expr rows
#' @param expr_cols  numeric vector: expr cols
#' @param fvar_rows  numeric vector: fvar rows
#' @param fvar_cols  numeric vector: fvar cols
#' @param svar_rows  numeric vector: svar rows
#' @param svar_cols  numeric vector: svar cols
#' @param fdata_rows numeric vector: fdata rows
#' @param fdata_cols numeric vector: fdata cols
#' @param sdata_rows numeric vector: sdata rows
#' @param sdata_cols numeric vector: sdata cols
#' @param transpose  TRUE or FALSE (default)
#' @param sfile      sample file
#' @param sfileby    sample file mergeby column
#' @param subgroupvar subgroupvar in sfile
#' @param verbose    TRUE (default) or FALSE
#' @return SummarizedExperiment
#' @examples
#' # RNASEQ
#'    file <- system.file('extdata/billing19.rnacounts.txt', package = 'autonomics')
#'    read_rectangles( file, fid_rows = 2:25,     fid_cols = 2,
#'                           sid_rows = 1,        sid_cols = 5:28,
#'                          expr_rows = 2:25 ,   expr_cols = 5:28,
#'                          fvar_rows = 1,       fvar_cols = 1:4,
#'                         fdata_rows = 2:25 ,  fdata_cols = 1:4,   transpose = FALSE)
#' # LCMSMS PROTEINGROUPS
#'    file <- system.file('extdata/billing19.proteingroups.txt', package = 'autonomics')
#'    read_rectangles(  file,
#'                      fid_rows = 2:21,    fid_cols = 383,
#'                      sid_rows = 1,       sid_cols = seq(124, 316, by = 6),
#'                     expr_rows = 2:21,   expr_cols = seq(124, 316, by = 6),
#'                     fvar_rows = 1,      fvar_cols = c(2, 6, 7, 383),
#'                    fdata_rows = 2:21,  fdata_cols = c(2, 6, 7, 383),
#'                    transpose  = FALSE )
#' # SOMASCAN
#'    file <- system.file('extdata/atkin.somascan.adat', package = 'autonomics')
#'    read_rectangles(file, fid_rows = 30,         fid_cols = 23:42,
#'                          sid_rows = 42:108,     sid_cols = 4,
#'                         expr_rows = 42:108,    expr_cols = 23:42,
#'                         fvar_rows = 28:40,     fvar_cols = 22,
#'                         svar_rows = 41,        svar_cols = 1:21,
#'                        fdata_rows = 28:40,    fdata_cols = 23:42,
#'                        sdata_rows = 42:108,   sdata_cols = 1:21,  transpose  = TRUE)
#' # METABOLON
#'    file <- system.file('extdata/atkin.metabolon.xlsx', package = 'autonomics')
#'    read_rectangles(file, sheet = 2,
#'                      fid_rows = 11:30,     fid_cols = 2,
#'                      sid_rows = 4,         sid_cols = 15:81,
#'                     expr_rows = 11:30,    expr_cols = 15:81,
#'                     fvar_rows = 10,       fvar_cols = 1:14,
#'                     svar_rows = 1:10,     svar_cols = 14,
#'                    fdata_rows = 11:30,   fdata_cols = 1:14,
#'                    sdata_rows = 1:10,    sdata_cols = 15:81,
#'                     transpose = FALSE )
#' @export
read_rectangles <- function(
    file, sheet = 1, fid_rows, fid_cols, sid_rows, sid_cols,
    expr_rows, expr_cols, fvar_rows  = NULL, fvar_cols = NULL, svar_rows = NULL,
    svar_cols  = NULL, fdata_rows = NULL,  fdata_cols = NULL, sdata_rows = NULL,
    sdata_cols = NULL, transpose  = FALSE,
    sfile = NULL, sfileby = NULL, subgroupvar = character(0),
    verbose = TRUE
){
    object <- .read_rectangles(file, sheet=sheet,
                        fid_rows   = fid_rows,   fid_cols   = fid_cols,
                        sid_rows   = sid_rows,   sid_cols   = sid_cols,
                        expr_rows  = expr_rows,  expr_cols  = expr_cols,
                        fvar_rows  = fvar_rows,  fvar_cols  = fvar_cols,
                        svar_rows  = svar_rows,  svar_cols  = svar_cols,
                        fdata_rows = fdata_rows, fdata_cols = fdata_cols,
                        sdata_rows = sdata_rows, sdata_cols = sdata_cols,
                        transpose  = transpose,  verbose    = verbose)
        object %<>% merge_sample_file(
            sfile = sfile, by.x = 'sample_id', by.y = sfileby, verbose=verbose)
    object
}



split_values <- function(x){
    sep <- guess_sep(x)
    dt <- data.table::data.table(x = x)
    dt[, data.table::tstrsplit(x, sep) ]
}




#' Merge sample/feature dt
#' @param object          SummarizedExperiment
#' @param dt              data.frame, data.table, DataFrame
#' @param by.x            string : object mergevar
#' @param by.y            string : df mergevar
#' @param all.x           TRUE / FALSE : whether to keep samples / features without annotation
#' @param verbose         TRUE / FALSE : whether to msg
#' @return                SummarizedExperiment
#' @examples
#' file <- system.file('extdata/atkin.metabolon.xlsx', package = 'autonomics')
#' object <- read_metabolon(file)
#' sdt(object)
#' sdt(merge_sdt(object, data.table(sample_id = object$sample_id,
#'                                     number = seq_along(object$sample_id))))
#'@export
merge_sdata <- function(
    object, dt, by.x = 'sample_id',  by.y = names(dt)[1], all.x = TRUE, verbose = TRUE
){
    .Deprecated('merge_sdt')
    merge_sdt(object, dt = dt, by.x = by.x, by.y = by.y, all.x = all.x, verbose = verbose)
}

#' @rdname merge_sdata
#' @export
merge_sdt <- function(
    object, dt, by.x = 'sample_id',  by.y = 'sample_id', all.x = TRUE, verbose = TRUE
){
    if (!all.x)  object %<>% filter_samples(!!sym(by.x) %in% unique(dt[[by.y]]), verbose = verbose)
    sdt(object) %<>% merge_data(dt, by.x = by.x, by.y = by.y, verbose = verbose)
    object
}


#'@rdname merge_sdata
#'@export
merge_fdata <- function(
    object, dt, by.x = 'feature_id', by.y = names(dt)[1], all.x = TRUE, verbose = TRUE
){
    .Deprecated('merge_fdt')
    merge_fdt(object, dt = dt, by.x = by.x, by.y = by.y, all.x = all.x, verbose = verbose)
}

#'@rdname merge_sdata
#'@export
merge_fdt <- function(
    object, dt, by.x = 'feature_id', by.y = 'feature_id', all.x = TRUE, verbose = TRUE
){
    if (!all.x)  object %<>% filter_features(!!sym(by.x) %in% unique(dt[[by.y]]), verbose = TRUE)
    fdt(object) %<>% merge_data(dt, by.x = by.x, by.y = by.y, verbose = verbose)
    object
}


merge_data <- function(objectdt, dt, by.x, by.y, fill = NULL, verbose){
# Assert
    if (is.null(dt))  return(objectdt)
    assert_is_data.table(objectdt)
    assert_is_data.table(dt)
    assert_is_subset(by.x, names(objectdt))
    assert_is_subset(by.y, names(dt))
# Prepare
    # Rm duplicate keys
        dt %<>% unique() # drop duplicate rows with identical info 
        n0 <- nrow(dt)   # drop rows with duplicate key values 
        dt %<>% unique(by = by.y) # keys should be unique!
        if (n0>nrow(dt) & verbose)  message('\t\t\tRetain ', nrow(dt),
            '/', n0, ' rows after removing duplicate `', by.y, '` entries')
    # Rm duplicate cols: https://stackoverflow.com/questions/9202413
        duplicate_cols <- setdiff(intersect(names(objectdt), names(dt)), by.x)
        for (dupcol in duplicate_cols) objectdt[, (dupcol) := NULL]
    # Ensure character class for merge column - one being factor means horrible bug!
    # Dont use dt[[byy]] : class not updated properly resulting in many NA when merging!
        for (byx in by.x)  objectdt[, (byx) := as.character(get(byx)) ]
        for (byy in by.y)        dt[, (byy) := as.character(get(byy)) ]
# Merge
    objectdt %<>% merge.data.table(dt, by.x = by.x, by.y = by.y, all.x = TRUE, sort = FALSE)
    objectdt
}


#' Merge sample / feature file
#'
#' @param object            SummarizedExperiment
#' @param sfile             string       : sample file path
#' @param ffile             string       : ffile path
#' @param by.x              string       : object mergevar
#' @param by.y              string       : file mergevvar
#' @param all.x             TRUE / FALSE : whether to keep samples / feature without annotation
#' @param select            character    : [sf]file columns to select
#' @param stringsAsFactors  TRUE / FALSE
#' @param verbose           TRUE / FALSE
#' @return SummarizedExperiment
#' @examples
#' file <- system.file('extdata/billing19.proteingroups.txt', package = 'autonomics')
#' subgroups <-  c('E00','E01', 'E02','E05','E15','E30', 'M00')
#' subgroups %<>% paste0('_STD')
#' object <- read_maxquant_proteingroups(file, subgroups = subgroups)
#' sfile <- paste0(tempdir(),'/', basename(tools::file_path_sans_ext(file)))
#' sfile %<>% paste0('.samples.txt')
#' dt <- data.table(sample_id = object$sample_id, 
#'                  day = split_extract_fixed(object$subgroup, '_', 1))
#' data.table::fwrite(dt, sfile)                 
#' sdt(object)
#' sdt(merge_sample_file(object, sfile))
#'@export
merge_sample_file <- function(
    object, sfile = NULL, by.x = 'sample_id', by.y = 'sample_id', all.x = TRUE, 
    select = NULL, stringsAsFactors = FALSE, verbose = TRUE
){
    if (is.null(sfile))  return(object)
    assert_all_are_existing_files(sfile)
    if (verbose)  message('\t\tMerge sdata: ', sfile)
    dt <- fread(sfile, select = select, stringsAsFactors = stringsAsFactors)
    assert_is_subset(by.y, names(dt))
    dt[[by.y]] %<>% as.character()
    object %<>% merge_sdt(dt, by.x = by.x, by.y = by.y, all.x = all.x, verbose = verbose)
    object
}

#' Merge sample excel
#' @param object SummarizedExperiment
#' @param sfile  sample file
#' @param range  string
#' @param by.x   string
#' @param by.y   string
#' @return SummarizedExperiment
#' @export
merge_sample_excel <- function(
    object, sfile, range = NULL, by.x = 'sample_id', by.y = 'sample_id'
){
# Assert
    assert_is_valid_sumexp(object)
    assert_all_are_existing_files(sfile)
    assert_is_subset(by.y, excelcols(sfile, range = range))
# Read, Merge, Return
    sdt0 <- read_excel(sfile, range = range)
    sdt0 %<>% data.table()
    sdt0$sample_id
    object %<>% merge_sdt(sdt0, by.x = by.x, by.y = by.y)
    object
}


#' @rdname merge_sample_file
#' @export
merge_ffile <- function(
    object, ffile = NULL, by.x = 'feature_id', by.y = 'feature_id', all.x = TRUE,
    select = NULL, stringsAsFactors = FALSE, verbose = TRUE
){
    if (is.null(ffile))  return(object)
    assert_all_are_existing_files(ffile)
    if (verbose) message('\t\tMerge fdata: ', ffile)
    dt <- fread(ffile, select = select, stringsAsFactors = stringsAsFactors)
    assert_is_subset(by.y, names(dt))
    dt[[by.y]] %<>% as.character()
    object %<>% merge_fdt(dt, by.x = by.x, by.y = by.y, all.x = all.x, verbose = verbose)
    object
}

add_subgroup <- function(
    object, subgroupvar = 'subgroup', verbose = TRUE
){
    # Rename
    object$subgroup <- object[[subgroupvar]]
    object[[subgroupvar]] <- NULL

    # Infer from sampleid if no subgroups defined yet
    if (!has_some_svalues(object, 'subgroup')){
        x <- object$sample_id
        sep <- guess_sep(x)
        if (sep!='NOSEP'){
            nfactor <- nfactors(x, sep)
            if (verbose)   cmessage('%sInfer subgroup from sample_ids', spaces(14))
            object$subgroup  <- split_extract_fixed(x, sep, seq_len(nfactor-1))
            object$replicate <- split_extract_fixed(x, sep, nfactor)
        } else {
            object$subgroup <- 'group0'
        }
    }
    
    # factorify/validify
    object$subgroup %<>% factor()
    levels(object$subgroup) %<>% make.names()  # otherwise limma issue!
    object
}

#==============================================================================
#
#                           read_affymetrix
#
#==============================================================================

# https://stackoverflow.com/a/4090208
# install_if_required <- function(pkgs){
#     pkgs %<>% extract(!(pkgs %in% installed.packages()[,"Package"]))
#     if(length(pkgs)) BiocManager::install(pkgs)
# }


add_affy_fdata <- function(object){
# Assert
    if (!requireNamespace('AnnotationDbi', quietly = TRUE)){
        message("`BiocManager::install('AnnotationDbi')`. Then re-run.")
        return(object)
    }
# Extract entrez identifiers
    entrezgs <- vapply(
        stri_split_fixed(fnames(object), '_'), extract, character(1), 1)
# Get annotation db
    pkgname <- paste0(metadata(object)$annotation, '.db')
    #install_if_required(pkgname)
    db <- getFromNamespace(pkgname, pkgname)
# Map
    rowData(object) <- DataFrame(
        feature_id    = fnames(object),
        feature_name  = suppressMessages(AnnotationDbi::mapIds(
                    db, entrezgs, column = 'SYMBOL',   keytype = 'ENTREZID')),
        feature_descr = suppressMessages(AnnotationDbi::mapIds(
                    db, entrezgs, column = 'GENENAME', keytype = 'ENTREZID')),
        row.names     = fnames(object)
    )
# Return
    object
}

#' Read affymetrix microarray
#' @param celfiles string vector: CEL file paths
#' @return RangedSummarizedExperiment
#' @examples
#' # Downloading example dataset fails 600s limit - example outcommented.
#' # url <- paste0('http://www.bioconductor.org/help/publications/2003/Chiaretti/chiaretti2/T33.tgz')
#' # localdir  <- file.path(tools::R_user_dir('autonomics', 'cache'), 'T33')
#' # dir.create(localdir, showWarnings = FALSE)
#' # localfile <- file.path(localdir, basename(url))
#' # if (!file.exists(localfile)){  download.file(url, destfile = localfile)
#' #                                untar(localfile, exdir = path.expand(localdir))  }
#' # localfile %<>% substr(1, nchar(.)-4)
#' # if (!requireNamespace("BiocManager", quietly = TRUE))  install.packages('BiocManager')
#' # if (!requireNamespace("hgu95av2.db", quietly = TRUE))  BiocManager::install('hgu95av2.db')
#' # read_affymetrix(celfiles = list.files(localfile, full.names = TRUE))
#' @export
read_affymetrix <- function(celfiles){
# Assert
    if (!requireNamespace('affy', quietly = TRUE)){
        stop("`BiocManager::install('affy')`. Then re-run.") }
# read
    message('Read Affymetrix CEL files: ', basename(celfiles)[1], ', ...')
    suppressWarnings(eset1 <- affy::just.rma(filenames = celfiles))
    object <- makeSummarizedExperimentFromExpressionSet(eset1)
# sdata
    snames(object) %<>% stri_replace_first_fixed('.CEL', '')
    sdata(object) <- data.frame(sample_id = snames(object),
                                row.names = snames(object),
                                stringsAsFactors = FALSE)
# fdata
    object %<>% add_affy_fdata()
# return
    return(object)
}


#==============================================================================
#
#                           read_genex
#
#==============================================================================


#' Read genex file
#' @param file string: path to exiqon genex file
#' @return SummarizedExperiment
#' @noRd
read_genex <- function(file){
    assert_all_are_existing_files(file)
    dt <- extract_rectangle(file, sheet=1)
    read_rectangles(
        file,
        sheet = 1,
        fid_rows   = 1,                     fid_cols   = 2:(ncol(dt)-2),
        sid_rows   = 2:(nrow(dt)-3),        sid_cols   = 1,
        expr_rows  = 2:(nrow(dt)-3),        expr_cols  = 2:(ncol(dt)-2),
        fvar_rows  = (nrow(dt)-2):nrow(dt), fvar_cols  = 1,
        svar_rows  = 1,                     svar_cols  = (ncol(dt)-1):ncol(dt),
        fdata_rows = (nrow(dt)-2):nrow(dt), fdata_cols = 2:(ncol(dt)-2),
        sdata_rows = 2:(nrow(dt)-3),        sdata_cols = (ncol(dt)-1):ncol(dt),
        transpose  = TRUE,
        verbose    = TRUE)
}


#' SummarizedExperiment list to long data.table
#' 
#' @param sumexplist  list of SummarizedExperiments
#' @param svars       character vector
#' @param fvars       character vector
#' @param setvarname  string
#' @return data.table
#' @examples
#' subgroups <- paste0(c('E00', 'E01', 'E02', 'E05', 'E15', 'E30', 'M00'), '_STD')
#' rnafile <- system.file('extdata/billing19.rnacounts.txt',     package = 'autonomics')
#' profile <- system.file('extdata/billing19.proteingroups.txt', package = 'autonomics')
#' fosfile <- system.file('extdata/billing19.phosphosites.txt',  package = 'autonomics')
#' rna <- read_rnaseq_counts(rnafile)
#' pro <- read_maxquant_proteingroups(file = profile, subgroups = subgroups)
#' fos <- read_maxquant_phosphosites(fosfile = fosfile, profile = profile, subgroups = subgroups)
#' pro$subgroup %<>% stringi::stri_replace_first_fixed('_STD', '')
#' fos$subgroup %<>% stringi::stri_replace_first_fixed('_STD', '')
#'     
#' sumexplist <- list(rna = rna, pro = pro, fos = fos)
#' dt <- sumexplist_to_longdt(sumexplist, setvarname = 'platform')
#' dt %<>% extract(gene %in% c('TNMD', 'TSPAN6'))
#' @export
sumexplist_to_longdt <- function(
    sumexplist, 
    svars = intersect('subgroup',  autonomics::svars(sumexplist[[1]])),
    fvars = intersect('gene',      autonomics::fvars(sumexplist[[1]])), 
    setvarname = 'set'
){
    assert_are_disjoint_sets(c(setvarname, 'xxxxx'), svars)
    assert_are_disjoint_sets(c(setvarname, 'xxxxx'), fvars)
    .sumexp_to_dt <- function(sumexp, set){
        dt <- sumexp_to_longdt(sumexp, svars={{svars}}, fvars={{fvars}})
        dt %<>% cbind(xxxxx = set)
        setnames(dt, 'xxxxx', setvarname)
    }
    rbindlist(mapply(.sumexp_to_dt, sumexplist, names(sumexplist), SIMPLIFY=FALSE))
}