more a more stable hash function

#' @export

#' @export

model.matrix.full <- function(vars_use, data_df, intercept = 0) {    

compute_hash <- function(data_df, vars_use) {

    discrete_vars <- c()

    base <- 1

    hash <- rep(0, nrow(data_df))

    for (varname in vars_use) {

    for (varname in vars_use) {

        if ("character" %in% class(data_df[[varname]])) {

        vals <- data.frame(data_df)[, varname, drop = TRUE]

            data_df[[varname]] <- factor(data_df[[varname]])

        nlevel <- nlevels(vals)

            discrete_vars <- c(discrete_vars, varname)

        hash <- hash + (as.integer(vals) - 1) * base

        } else if ("factor" %in% class(data_df[[varname]])) {

        base <- base * nlevel

            discrete_vars <- c(discrete_vars, varname)            

        } else {

            stop('no support yet for continuous variables')

    }

    }

    }    

    return(hash)

    contrasts_df <- lapply(discrete_vars, function(x) {

        diag(nlevels(data.frame(data_df)[, x, drop = TRUE]))

    })

    names(contrasts_df) <- discrete_vars

    for (varname in discrete_vars) {

        colnames(contrasts_df[[varname]]) <- levels(data_df[[varname]])

        row.names(contrasts_df[[varname]]) <- levels(data_df[[varname]])

    }

    res <- model.matrix(as.formula(sprintf("~ %d + %s", intercept, paste(vars_use, collapse = "+"))), 

                 data=data_df, contrasts.arg=contrasts_df)    

    return(res)

}

}

#' @export

#' @export

collapse_counts <- function(counts_mat, meta_data, varnames) {

collapse_counts <- function(counts_mat, meta_data, varnames) {

    design <- model.matrix.full(varnames, meta_data)

    ## give each unique row a hash value for indexing

    ## give each unique row a hash value for indexing

    hash <- factor(design %*% matrix(2 ^ seq(0, ncol(design) - 1), ncol = 1))

    hash <- compute_hash(meta_data, varnames)

    idx_keep <- which(!is.na(hash))

    hash <- hash[idx_keep]

    hash <- factor(sprintf('sample_%d', as.integer(hash)))

    hash <- factor(sprintf('sample_%d', as.integer(hash)))

    length(unique(hash))

    meta_data <- meta_data[idx_keep, ]

    counts_mat <- counts_mat[, idx_keep]

    ## one hot encoded design matrix, sample level

    ## one hot encoded design matrix, sample level

    design_collapsed <- design %>% 

    design_collapsed <- data.frame(meta_data)[, varnames] %>% 

        as_tibble() %>% 

        cbind(sample_id = hash) %>% 

        cbind(sample_id = hash) %>% 

        unique()

        unique()

    row.names(design_collapsed) <- design_collapsed$sample_id

    row.names(design_collapsed) <- design_collapsed$sample_id

    ## in case cells were dropped b/c of NA values

    ## sum over samples

    counts_mat <- counts_mat[, row.names(design)]

    counts_collapsed <- presto:::sumGroups(counts_mat, hash, 1) %>% t()

    counts_collapsed <- presto:::sumGroups(counts_mat, hash, 1) %>% t()

    row.names(counts_collapsed) <- row.names(counts_mat)

    row.names(counts_collapsed) <- row.names(counts_mat)

    colnames(counts_collapsed) <- levels(hash)

    colnames(counts_collapsed) <- levels(hash)

    ## reorder to match design matrix

    ## reorder to match design matrix

    counts_collapsed <- counts_collapsed[, design_collapsed$sample_id]

    counts_collapsed <- counts_collapsed[, design_collapsed$sample_id]

    ## recover the 

    return(list(counts_mat = counts_collapsed, meta_data = design_collapsed))

    meta_collapsed <- data.frame(row.names = row.names(design_collapsed))

    for (varname in varnames) {

        matching_colnames <- grep(sprintf('^%s', varname), colnames(design_collapsed), value = TRUE)

        vals <- matching_colnames[apply(design_collapsed[, matching_colnames] == 1, 1, which)]

        meta_collapsed[[varname]] <- gsub(varname, '', vals)

}

}

    return(list(counts_mat = counts_collapsed, meta_data = meta_collapsed))

}

#' @export

#' @export

pseudobulk_deseq2 <- function(dge_formula, meta_data, counts_df, verbose=TRUE, 

pseudobulk_deseq2 <- function(dge_formula, meta_data, counts_df, verbose=TRUE, 

                   min_counts_per_sample=10, present_in_min_samples=5) {

                   min_counts_per_sample=10, present_in_min_samples=5) {