<table border="1" cellspacing="0" cellpadding="8">
    <tr>
        <th>Issue</th>
        <td>
            <a href=https://github.com/llvm/llvm-project/issues/107477>107477</a>
        </td>
    </tr>

    <tr>
        <th>Summary</th>
        <td>
            [mlir][sparse] Adding two COO tensors and returning result fails with Python bindings
        </td>
    </tr>

    <tr>
      <th>Labels</th>
      <td>
            mlir
      </td>
    </tr>

    <tr>
      <th>Assignees</th>
      <td>
      </td>
    </tr>

    <tr>
      <th>Reporter</th>
      <td>
          mtsokol
      </td>
    </tr>
</table>

<pre>
    Hi!

I've been writing a short program with MLIR and Python bindings and I noticed that adding two COO tensors and returning COO result throws an error:
```
terminate called after throwing an instance of 'std::length_error'
  what():  vector::_M_fill_insert
Aborted (core dumped)
```

**I'm using LLVM** `19.1.0-rc3`

---

My setup is:
1. I create NumPy arrays `pos` (of dtype `int64`), `index` (of dtype `int64`), and `data` (of dtype `float64`) and pass them to MLIR module that assembles them and returns a pointer to the COO tensor
2. Then I pass that pointer to another MLIR module that adds COO tensors element-wise `add(a, a)` and returns the result as a pointer to the resulting COO.

It fails with an error pasted above. Here's a full python script that reproduces it (less than 100 lines): https://gist.github.com/mtsokol/a1b83817c02353242ae6be4fae10eb95

---

I did spend some time investigating it and I think the culprit is with returning the result COO tensor after addition. If I convert the COO tensor to a dense one before returning (so step 3. let's say) the program works: https://gist.github.com/mtsokol/06e0f5421b190aaa4eed0dcf560dd3bb

Also, it shouldn't be the problem with the way I return the COO tensor. To prove it here's one more gist (mostly copied from previous ones) where I assemble a COO tensor, return a pointer to it and pass it to another module call "disassemble" and return constituent NumPy arrays and successfully compare them with original ones: https://gist.github.com/mtsokol/d4b4c16a440de03f700601e91cd7f745

My suspicion the issue is with the returning COO tensor that is the result of addition directly from the call, which I would classify it as a bug.
</pre>
<img width="1px" height="1px" alt="" src="http://email.email.llvm.org/o/eJyUVtFu4zoO_Rrlhaghy46dPOShM0UxAaY7g8VgXwvZomPtyJIh0cnm7xe03TYtinvvAIEDSyTFQx4eS6dkTx7xILZfxPZhoyfqQzwMlMLv4DZNMNfDNytULuSDkPfL8yhUfUZoED1coiXrT6Ah9SESjDGcoh7gYqmHp-_Hf4P2Bn5eqQ8eGuuN9ac0rx3BB7ItGqBeE2jDW0CXAF9__ABCn0JcLCPSFD3v8k7ENDkC6mO48D5gjCGKYk1OVHL9za-EcbBeE0KrnUMDuiOMi_ectwfrE2nfIoQOhKoTGQ5W3Dv0J-qfl_CqXuIBXHpNQu2E2oviHuCMLS3HF_fPT8-dde7Z-oSRFof7JkRCA0Lt2hARzDSMaNj7s3TXp-Ifl3mAKXGa37__52lZBVHJfJ_lmbyLbfHB8e7u7vb16QoJaRrBptf65BkcoY3IJfnXNPy8go5RXxPHHUMSleRUQweGriPyqvVUlXyQ2gv1dVkx-L-_t-TeiUoaTfoT484F_Wo-2446JaAeB6CwcGcIZnK4EiQlHBqHq8kbMRJoGIP1c18D794waAGtMvjVo4fjyxmabl20D9Rj_ORMY9I7OqLDAT3dXWyaQWhjhNrpGS2jruS7xDiXla_6kzSXrZXY2bsRI-i0dWmZoxeSc_bMJd2EM2bwDSMyYUFDNzkH4zJlqY12pAVAxDEGM7WYwBJ3wOGC30MuJTjrMa1M7onGmSbqUajHk02UnSz1U5O1YRDqcdUEoR513uyKXV63UhXbQpVKY9Vg2WnMJTb77V8w8gjGGkgjegMpDAhkBwTrz5jInvRcC0urPlBv_e-5UO3kxmgJ7FqQN0W4qfBbn9YhZ0khG3wGx45JH_wZI30gyNx_MOgTQvCsah2P6dsJQu1SgEQ4QpGBQ5pLnvSVacuxXiUvxN_pzyopK5TdtlR5k--l1rpENNK03baSxhRNc1u6e5cC88wSS-3kjBeqJmjwJYnG4aq7vHDRVziuMD5AzuBXYIczcrD-hUWMfmDsnDHDHkIid4U2jBYNdDEMMEY82zDNxkwcuLA7HF-nE_Tt8KmvLxm84_7a4XkWLd1O4Dp8rNUglDI2vQQWSt2MFjczkaUJPb2XMbZJU9tiSjwUnP4w6oiLaszlCdGerNduAfFHDTNlU7Z5pctSGpRFV0tZyRz3eWvqri63H9V3SqNtbVg6YFOa8JXDC3Nvv2wvhOTBte_EI3SvZAZjI7bcl7kh83Ro57jUl962PRzhwuyA1vHHvbvO1WaNaKZTtjGHwuyLvd7gIa_VdluXu12-6Q-yULKqNBZ10-gCy2rX5Fi3dSmxMzutN_agpCrlXm6VUrnKMyz2-6JWed20VbfTnSglDtq6zLnzkIV42sx4D7msy7reON2gS_M9Q6nB2SiU4htHPLD9XTOdkiils4nSWwSy5Oa7yeywfRDbL2nUMaHYPsD9P7kxrAW8EdMPd5HNFN3hIwFues-5rH93Ywz_xZaEepyhJaEeV3Tng_p_AAAA__-lcAoY">