[AgustinPardo]

Handle positions with zero accumulated probability when generating the decoded sequence and quality string

• Avoid division by zero when totalProbs[i] is zero
• Prevent log10(0) by clamping the error probability to the smallest positive float
• Preserve the existing quality-score scaling and clamping behavior

[malton-ont]

Hi @AgustinPardo,

Thanks for your contribution.

Have you raised this because you've had an issue caused by this code (I don't see an open ticket)? I ask because this change introduces additional branching into a critical loop, which could have a deleterious effect on performance. As far as I can see it's not possible for the most probable base to have a base_probs of 0 and total_probs should sum to 1 so that can't be 0 either. Therefore this change seems to be redundant.

If you're experiencing something different, we'd like to take a look and solve the issue at source rather than masking it with this change.

[AgustinPardo]

Hi @malton-ont

Thanks for the quick review and the detailed explanation.

You're right , I raised this as a precautionary fix based on static code analysis rather than an observed failure in practice. I didn't have a concrete reproduction case, and I should have opened a ticket first to discuss whether the scenario is actually reachable before submitting the PR.

Your explanation makes sense: if totalProbs always sums to 1 by construction, then the guard is indeed redundant and the added branching in a hot loop is a valid concern.