Some ARAX results are being dropped during the "result transfomer" stage of processing

[dkoslicki]

Compare: https://arax.ci.transltr.io/?r=427688
with: https://arax.transltr.io/?r=427689
Note that the CI result says there are 3 results, but truncates it to 1 for some reason. Prod says there are three and returns three.

Tags say that prod was last updated at May 7th, 2025. Looks like in July, #2525 touched resultify, so tagging Steve

[saramsey]

Going to work on this issue in branch issue-2620

[saramsey]

Here is the query graph that had been sent to CI, that produced result 427688.

{
  "edges": {
    "t_edge": {
      "attribute_constraints": [],
      "knowledge_type": "inferred",
      "object": "ON",
      "predicates": [
        "biolink:treats"
      ],
      "qualifier_constraints": [],
      "subject": "SN"
    }
  },
  "nodes": {
    "ON": {
      "categories": [
        "biolink:Disease"
      ],
      "constraints": [],
      "ids": [
        "MONDO:0004260"
      ],
      "is_set": false
    },
    "SN": {
      "categories": [
        "biolink:ChemicalEntity"
      ],
      "constraints": [],
      "is_set": false
    }
  }
}

Here is the query graph that had been sent to Translator PROD, that produced result 427689, for comparison:

{
  "edges": {
    "t_edge": {
      "attribute_constraints": [],
      "knowledge_type": "inferred",
      "object": "ON",
      "predicates": [
        "biolink:treats"
      ],
      "qualifier_constraints": [],
      "subject": "SN"
    }
  },
  "nodes": {
    "ON": {
      "categories": [
        "biolink:Disease"
      ],
      "constraints": [],
      "ids": [
        "MONDO:0004260"
      ],
      "is_set": false,
      "set_interpretation": "BATCH"
    },
    "SN": {
      "categories": [
        "biolink:ChemicalEntity"
      ],
      "constraints": [],
      "is_set": false,
      "set_interpretation": "BATCH"
    }
  }
}

[saramsey]

There does seem to be a difference between the two query graphs, in that the QG shown that came from CI does not have "set_interpretation": "BATCH" for the nodes, and the QG shown that came from Translator PROD does.

[saramsey]

In the results from running the QG on arax.ci.transltr.io, hyperlinked by @dkoslicki as ARAX result number 427688, the knowledge graph seems to have three nodes that are linked to MONDO:0004260 ("peptic ulcer perforation") by a biolink:treats edge:

• cimetidine (CHEBI:3699)
• teicoplanin (CHEBI:29687)
• thymalfasin (CHEBI:135915)

the three nodes are shown here in the KG network diagram in the ARAX web browser UI:

Now, in result 427688, only "cimtidine" makes it into the "Results" tab:

and it is interesting that the UI displays the message "truncated from a total of 3" (I will come back to that in a moment).

What is unique about "cimetidine", that might explain why it is shown as a result and "teicoplanin" and "thymalfasin" are not? Well, the cimetidine-[treats]->peptic_ulcer_perforation edge is supported by a lookup type triple from MolePro, as shown here:

whereas the other two results seem to be creative-mode results, as shown here for teicoplanin:

and here for thymalfasin:

[saramsey]

In the results from the query that @dkoslicki ran on the Translator PROD arax (arax.transltr.io), i.e., result number 427689, displayed using the ARAX web browser interface on arax.transltr.io, we see all three results, and there is no "truncated from a total of 3" text label displayed in the ARAX web browser UI:

What if I display the same result JSON (427689), but change the ARAX web browser UI server to arax.ci.trnsltr.io? It still shows the three results, without the "truncated from a total of 3" message:

And what if I diplay the result JSON (427688) from the query being run on CI, and display it using the ARAX web browser UI server arax.transltr.io? Will it make any difference? No, it doesn't change the outcome; the UI displays one result and it displays the message "truncated from a total of 3":

[saramsey]

To me, the results from this reciprocal test seem to suggest that the cause, whatever it is, is not a change in the ARAX web browser UI between the last ITRB PROD version of ARAX and the current CI version of ARAX. I need to dig further to understand where this "truncated from a total of __" text originates in the ARAX code.

[saramsey]

It appears that the "truncated from a total of ___" text label is created in the ARAX web browser interface code (rtx.js):

RTX/code/UI/interactive/rtx.js

Line 1970 in bf4b0c3

rtext += " (truncated from a total of " + respObj.total_results_count + ")";

but of course, that doesn't necessarily mean that the fault is in the rtx.js module; the code for creating and appending that text label is in an if conditional. I suspect that something in result 427688 is triggering the conditional, whereas the corresponding datum in result 427689 is not triggering the conditional.

Here is the full if block:

	    if (respObj.total_results_count && respObj.total_results_count > respObj.message.results.length)
		rtext += " (truncated from a total of " + respObj.total_results_count + ")";

[saramsey]

So, at L1969 in rtx.js, evidently .length retrieves a stored value from a Javascript object. I need to find where this value is set in the object.

[isbluis]

Hi @saramsey . The x.length is just the size of the array in Javascript, in this case the number of results.
And as you found out (faster than I could respond!), the UI will show that message when the total_results_count (a property of the result object sent by the backend) is larger than the actual number of results sent back.

[saramsey]

This is helpful, thanks @isbluis

[saramsey]

I downloaded the result JSON for result 427689, and indeed, it has three results, as shown here:

and the result JSON for results 427688 indeed has one result, as shown here:

To me, this is pretty conclusive evidence that the difference is arising in on of the the ARAX server modules and not in the ARAX web browser interface.