Use `patternProperties` to match columns with a certain `prefix`

[agusinac]

In my samplesheet.csv I wish to include other columns, such as CONTRAST_ or VARIABLE_ that will be used downstream in post-analysis.

In my R package, I use something similar as follows, that does perform pattern matching. Unfortunately, this is not supported and the samplesheetToList ignores all patternProperties despite when additionalProperties is set to true.

"patternProperties": {
          "^CONTRAST_": {
            "anyOf": [
                { "type": "string", "pattern": "^(\\S*)$" },
                { "type": "number" }
              ]
            },
          "^VARIABLE_": {
            "anyOf": [
              { "type": "string", "pattern": "^(\\S*)$" },
              { "type": "number" }
            ]
          }
        },

At the moment I made my own samplesheetToMetadata function that simply takes everything and returns it as a list. Of course, this is not the ideal approach..

def samplesheetToMetadata(input) {
    def rows = []
    input.withReader { reader ->
        def headers = reader.readLine().split(',').collect { it.trim() }
        reader.eachLine { line ->
            def values = line.split(',').collect { it.trim() }
            def row = [:]
            headers.eachWithIndex { h, i -> row[h] = values[i] }
            rows << row
        }
    }

    def isNumeric = { str ->
        str ==~ /^-?\d+(\.\d+)?$/
    }

    // Check types for each column
    def columnTypes = [:]
    if (rows) {
        def headers = rows[0].keySet()
        headers.each { col ->
            def values = rows.collect { it[col] }
            def allNumeric = values.every { v -> isNumeric(v) }
            def allString = values.every { v -> v instanceof String && !isNumeric(v) }
            columnTypes[col] = allNumeric ? 'numeric' : (allString ? 'string' : 'mixed')
        }
    }

    return [rows]
}

I didn't bother looking into samplesheetToList because I needed a quick fix to move on, but I might look into this myself soon if you nobody else is looking into it.

Also, on another note, I noticed that when I want to use the same schema_input.json in R.. it doesn't follow the correct json syntax according to json validator