#Reading a table

#Getting the table

The deck ships no real data. The published frequencies belong to their authors, and STRidER's terms of use reserve redistribution:

The reproduction, duplication, editing, dissemination and all type of utilization beyond the copyright require written approval of the respective author or producer. Downloads and copies of these pages are only for private and not for commercial use.

Download the XML yourself from https://strider.online/frequencies/xml. That permission covers private, non-commercial use; anything beyond it is a question for the rights holder. testdata/frequencies.xml is synthetic and carries no STRidER content.

use fs;
def xml as string init fs.readString("strider.xml");

#One export, many populations

A download holds 33 tables in one document: the whole database, each continent and each contributing country. parse reads them all; you choose the population afterwards by name:

def t as strider.Table init strider.parse($xml);
strider.populations($t);   # ["Entire Database", "Africa", ..., "AUSTRIA", ...]

A country with few samples carries fewer loci than the whole database, so loci is asked per population.

#Parse, check, convert

def sums as map of string to float init strider.frequencySums($t, "AUSTRIA");
for (def locus in $sums) {
    if (math.abs($sums[$locus] - 1.0) > 0.01) {
        io.printf("locus %s sums to %f, not 1.0\n", $locus, $sums[$locus]);
    }
}

def db as forensics.FrequencyDb init
    strider.toFrequencyDb($t, "AUSTRIA", "STRidER AT");

A locus far from 1.0 was read wrongly: a truncated download, a percentage-scaled table, or the wrong value taken as the frequency. Conversion rejects a locus summing above 1.05, but a locus summing to 0.6 converts without complaint; only this check catches it.

Use the bulk sum accessors, not a loop over frequencySum. A Table is a value, so each accessor call copies the whole export, every population included. Over a real download, the same 704 sums cost:

allFrequencySums($t) audits the whole export, frequencySums the population you will compute with, and frequencySum a single locus.

load parses and converts in one call:

def db as forensics.FrequencyDb init strider.load($xml, "AUSTRIA", "STRidER AT");

It skips the check, and calling it per population parses the whole document (about half a megabyte) each time. For more than one population, parse once and call toFrequencyDb for each.

#The shape it reads

Locus-major, with the populations nested inside each locus (numbers illustrative):

<frequencies created_timestamp="..." validity="...">
  <marker>
    <name>D3S1358</name>
    <alleles>15, 16, 17</alleles>
    <origin name="Entire Database" n="20000">
      <frequency allele="15">2.50000e-1</frequency>
    </origin>
    <origin name="AUSTRIA" n="500">
      <frequency allele="15">3.00000e-1</frequency>
    </origin>
  </marker>
</frequencies>

Tag and attribute names match in any case, and markers are found at any depth. The <alleles> list is ignored. A marker with no <origin> is a locus typed in no population; it contributes nothing. The real export has one.

#Frequencies are kept as published

No floor and no renormalisation: the Table holds the document's numbers, so frequencySum is meaningful. The NRC II 5/(2N) floor is applied on read by forensics.alleleFrequency, which is why a FrequencyDb keeps a per-locus N.