My last post was AI letter grades. This one strips the comparison down to two steps — search and replace — with actual seconds.
Two caveats: the data comes from two machines (a 16 GB Windows laptop and a 32 GB M4 Mac), so tables are kept separate and cross-environment comparison is off the table. And search terms / run counts / replace methods vary slightly per source article, so only same-table comparisons are valid. I develop UwView; every number is my own measurement. (Size note: the largest target is one and the same file — ~892M lines — written as 50 GB / 51.25 GB / 47.73 GiB across my source posts; here it's standardized to 48 GB.)
Windows (Win11, 16 GB RAM, internal SSD)
Replace — two replacements (東京→Tokyo, 大阪→Osaka):
| Size | EmEditor | UVP+Edit | Gap |
|---|---|---|---|
| 3 GB / 100M lines | 0.9 s | 6.3 s | EmEditor 7.0× |
| 10 GB / 100M lines | 28.6 s | 12.41 s | UVP+Edit ~2.3× |
| 48 GB / 892M lines | 672.2 s* | 47.0 s | UVP+Edit ~14.3× |
* The 2nd replacement hung (unfinished at 13 min); 672.2 s uses the value re-measured after a reboot.
Search — full-text 東京 on 48 GB: EmEditor 160 s (single run) / 194.7 s (two-run avg) vs 12.5 s / 23.9 s for the UwView family (~8–12.8×). Standalone 3/10 GB search wasn't measured on Windows.
Mac (MacBook Air M4, 32 GB RAM)
Search (GUI) vs klogg — 10-word sequential search, three storage types:
| Size | klogg | UwView Pro | Ratio |
|---|---|---|---|
| 3 GB | ~2 s | 0.40–0.72 s | ~3.9× |
| 10 GB | ~2–3 s | 1.14–2.41 s | ~1.7× |
| 48 GB | 17.5–585 s | 5.1–74.8 s | ~7.7× |
At 3–10 GB klogg answers in seconds too — small practical gap. Past RAM, klogg re-reads the disk per search; UwView Pro reads a ~1/9-size compressed cache.
Search (CLI) vs grep/ripgrep — one search over the same 48 GB file on USB storage (0.41 GB/s, I/O-bound): grep 64.64 s, rg 71.49 s, UwView Pro 54.7 s first run — then 14 s on repeats. One-off? Use grep. It even beats rg here (parallelism doesn't pay on one huge file).
Search & replace vs 010 Editor / UltraEdit (search 東京, replace-all → Tokyo): at 10.19 GB, 010 Editor 11.5 s / 16.5 s (re-measured 2026-09-08) vs 2.2 s / 3.6 s; at 3.07 GB, UltraEdit 5.3 s / 14.1 s vs 0.94 s / 1.31 s. At ~48 GB, UltraEdit is not measurable (it warns above 5 GB); 010 Editor opens in 56.7 s and completes one replace in 12 m 45.3 s — but its save crashed after 15+ minutes.
Summary
- Replace: EmEditor 7.0× faster at 3 GB; flips at 10 GB; 14.3× the other way at 48 GB
- Search: UwView family leads at every size, but within RAM everyone answers in seconds — past RAM, only the other tools drop into minutes
- Exception: the one-off CLI search — grep, simplest and fastest
Full disclosure
- No cross-environment comparisons; some cells are simply unmeasured
- All numbers are my own, on specific hardware and files — not an independent benchmark. Rerun anything and I'll update to match
UwView Pro is on sale (Windows, macOS, Linux — one license covers all three; 14-day free trial — editing features included). Full tables, per-measurement conditions, and links to all source posts:
https://uvp.y42u.net/en/blog/uvp-search-replace-3-10-50gb-en/
Top comments (0)