You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Public attention metrics can emit Infinity or convert booleans into measurements #782
On current main bcd699f, fetch_attention_feeds admits positive Infinity and booleans as numeric metrics. A public JSON number such as 1e400 decodes to Infinity in Python; serializing the resulting observation with allow_nan=False fails. A boolean becomes an invented measurement of 1.0. An oversized integer can instead discard the entire feed through an overflow exception.
The same comprehensions occur in primary observations, supporting observations, and restored observations after a Hacker News collector failure. Native caller regressions fail on all three paths. Measured zero and ordinary nonnegative numeric metrics are valid controls.
Proposed change
Use one finite, nonnegative numeric normalizer across those paths, excluding booleans and invalid metric containers. Preserve the observation and its valid measurements rather than fabricating a value or emitting invalid JSON.
This is separate from #616's new leaderboard ranking signals and #704's source-connector missing-counter handling. It changes public-attention feed normalization only. No committed source observations or generated artifacts will be edited.
Closing as not planned. One author now has a cap of 10 open issues + PRs, and this account had 56 open PRs and 11 open issues, mostly paired with each other. The paired PR is closed too. We merged the three changes that fix silent data loss in the corpus and are closing the rest so review stays possible.
This is not a verdict that the report is wrong. If you think it matters, reopen it later when you are under the cap, with a reproduction from a real user or caller rather than a hypothetical edge case.
Activity
ktwu01 commentedon Oct 7, 2026
Closing as not planned. One author now has a cap of 10 open issues + PRs, and this account had 56 open PRs and 11 open issues, mostly paired with each other. The paired PR is closed too. We merged the three changes that fix silent data loss in the corpus and are closing the rest so review stays possible.
This is not a verdict that the report is wrong. If you think it matters, reopen it later when you are under the cap, with a reproduction from a real user or caller rather than a hypothetical edge case.
不是否定这个问题本身。为保证评审能跟上,同一作者最多同时保留 10 个未关闭的 issue/PR。对应的 PR 也已关闭。若认为重要,请在低于上限后重新打开,并附真实用户或调用方的复现。
Generated by Claude Code