Conditional ElasticSearch sorting by different fields if in range - sorting

Question

I do have have products and some of them are reduced in price for a specific date range.
(simplified) example products:
{
"id": 1,
"price": 2.0,
"specialPrice": {
"fromDate": null,
"tillDate": null,
"value": 0,
},
},
{
"id": 2,
"price": 4.0,
"specialPrice": {
"fromDate": 1540332000,
"tillDate": 1571781600,
"value": 2.5,
},
},
{
"id": 3,
"price": 3.0,
"specialPrice": {
"fromDate": null,
"tillDate": null,
"value": 0,
},
}
Filtering by price was no problem. That I could do with a simple bool query.
But I could not yet find a good example for ElasticSearch scripts that could point me in the right direction, even though it should be quite simple, given you know the syntax.
My pseudocode: price = ('now' between specialPrice.fromDate and specialPrice.tillDate) ? specialPrice.value : price
Is there a way to translate this into something that would work in an ElasticSearch sorting?
To clarify further: By default, all products are already sorted by several conditions. The user can also search for any terms and filter the results while also being able to select multiple sorting parameters. Items can for example be sorted by tags and then by price, it's all very dynamic and it does still sort those results by some other properties (including the _score) afterwards.
So just changing the _score would be bad, since that is already calculated in a complex matter to show the best results for the given search terms.
Here is my current script, which does fail at the first params.currentDate:
"sort": {
"_script": {
"type": "number",
"script": {
"source": "if(doc['specialPrice.tillDate'] > params.currentDate) {params.currentPrice = doc['specialPrice.value']} return params.currentPrice",
"params": {
"currentDate": "now",
"currentPrice": "doc['price']"
}
}
}
How it does work now:
One problem was the nesting of some of the properties.
So one of my steps was to duplicate their content to new fields for the product (which I'm not that happy about, but whatever).
So in my mapping, I created new properties for products (specialFrom, specialTill, specialValue) and gave the corresponding fields in my specialPrice "copy_to" properties with the new property names.
The part is in php array syntax, since I'm using ruflin/elastica:
'specialPrice' => [
'type' => 'nested',
'properties' => [
'fromDate' => [
'type' => 'date',
'format' => 'epoch_second',
'copy_to' => 'specialFrom',
],
'tillDate' => [
'type' => 'date',
'format' => 'epoch_second',
'copy_to' => 'specialTill',
],
'value' => [
'type' => 'float',
'copy_to' => 'specialValue',
],
],
],
'specialFrom' => [
'type' => 'date',
'format' => 'epoch_second',
],
'specialTill' => [
'type' => 'date',
'format' => 'epoch_second',
],
'specialValue' => [
'type' => 'float',
],
Now my sorting sorting script does look like this (in my testing client, still working on implementing it within elastica):
"sort": {
"_script": {
"type": "number",
"script": {
"lang": "painless",
"source": "params.param = ((doc['specialTill'].value - new Date().getTime()) > 0 && (new Date().getTime() - doc['specialFrom'].value) > 0) ? doc['specialValue'].value : doc['price'].value; return params.param;",
"params": {
"param": 0.0
}
}
}
}
I'm not 100% happy with this because I have redundant data and scripts (calling new Date().getTime() twice in the script), but it does work and that is the most important thing for now :)

Answer 1

I've updated the below query post your clarifications. Let me know if that works!
POST dateindex/_search
{
"query":{
"match_all":{ // you can ignore this, I used this to test at my end
}
},
"sort":{
"_script":{
"type":"number",
"script":{
"lang":"painless",
"inline":" params.param = ((doc['specialPrice.tillDate'].value - new Date().getTime()) > 0) ? doc['specialPrice.value'].value : doc['price'].value; return params.param;",
"params":{
"param":0.0
}
},
"order":"asc"
}
}
}
You can try using source instead of inline in the above query as I've been testing on ES5.X version on my machine.
Hope it helps!