Automattic / Automattic/harper
Harper [Obsidian] stops detecting anything after a file reaches a certain length (I'm not sure how long exactly)
- Dominant language
- Rust
- Stars
- 15.4k
- Forks
- 627
- Avg merge
- 1d 12h
- Merged PRs (30d)
- 106
Description
**Describe the bug**
Harper stops noticing errors in long Obsidian files, specifically after some sort of cutoff point. It can still detect errors earlier on in the same file, but not further down/where I'm currently working.
**To Reproduce**
Steps to reproduce the behavior:
1. Write in the same Obsidian file for a long time. Currently the file I'm having the issue with is 10,672 words long (71,491 characters).
2. At some point, I'm not sure exactly where, Harper doesn't detect errors. Even if I write a string of nonsense ("fjakldfjl;adjfds") or a nonsense word which is detected in other files ("fark"), it isn't flagged.
**Expected behavior**
I expect Harper to continue detecting errors when I have plenty of available system resources. I'm not sure how long a file must be for it to stop doing its thing or why it happens.
**Screenshots**
Unfortunately, the note I'm working in contains confidential/copyrighted work which I am not at liberty to share, therefore, I can't take a screenshot that illustrates the issue.
**Platform**
I'm using Obsidian for macOS.
**Additional context**
The plugin is up to date and I've already tried reloading/launching Obsidian and toggling Harper on/off.
Hardware: MacBook Air M4 (13"), 16GB RAM, 256GB SSD
Software: latest Obsidian, macOS Tahoe 26.5.1
Contributor guide
Research direction
Start by reproducing the issue in the Harper Obsidian plugin on macOS with a long note, using the reported 10,672-word and 71,491-character file sizes as reference. Check whether errors are detected near the beginning but not near the end, and determine the length or condition that triggers the cutoff. Done means long Obsidian files continue reporting errors throughout, with a regression test or documented reproduction covering the failure.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- macos, rust
- Domain
- desktop
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 38/100