Update (17 Sep 2026): I thought we lost at 3 GB — turns out we’d won (as long as the cache is cold)
This article says, in one form or another, that ripgrep is faster at 3 GB. That was true while the file was in the OS page cache. Measured straight after
sudo purge— that is, opening a file you have not touched for a while — uvp is 2.6× faster even at 3 GB.
3 GB, fixed string rg uvp Cold (right after purge) 3.26 s 1.25 s (2.6×) Warm (again immediately) 0.33 s 1.01 s Six of seven search types behave the same way (uvp 1.6.3). The detail is in I thought we lost at 3 GB — turns out we’d won. Measured on a Mac (Apple M4 / 32 GB / external USB SSD); Windows and Linux — a low-powered laptop, under VMware — may behave differently. The figures below are unchanged (they were measured warm).
This is the run log of uv_cli_test4.sh, which checks every uvp command from the blog articles against rg / sort / uniq / sed / head / tail / gzip on real 3 GB, 10 GB and 50 GB files. What the commands mean and how to use them is covered in We shipped a uvp command and the CLI-to-UwView-Pro cheat sheet, so it is not repeated here.
Run log (complete, unedited)
% ./uv_cli_test4.sh
# uvp commands from the articles, checked against the CLI (rg / sort / uniq / sed) — run 20260915-170224-uvp4
| Item | Value |
|---|---|
| OS | Darwin 25.3.0 arm64 |
| Machine | Apple M4 / 32GB |
| uvp | /usr/local/bin/uvp uvp 1.6.0 |
| ripgrep | /opt/homebrew/bin/rg (ripgrep 15.2.0) |
| sed | /usr/bin/sed (BSD sed) |
| Data | /Volumes/BIWIN/26work/UwViewData/osm |
| .uwvz | existing sidecar reused (built first if missing) |
| Hit limit | counted across all shards; rows whose first term can exceed 1,000,000 hits carry `-limit none` |
| Checks | search icase regex drill invert grepform context seq tally sortcnt sortval headtail sorthead outtxt outgz sedline uwvzonly extract cat exit1 regexnl exit2 limitn limitnone replace replacere |
== 3G: osm/japan-dv-ai 3032812644 bytes
.uwvz present (251829224 bytes) → reusing it
$ rg -n -F '東京' osm/japan-dv-ai
$ uvp osm/japan-dv-ai '東京'
✅ search (literal) CLI 3.31 s uvp 0.79 s CLI/uvp=4.19 🟢 11274 hits exit=0 .uwvz reused
$ rg -n -i -F tokyo osm/japan-dv-ai
$ uvp osm/japan-dv-ai tokyo -i
✅ ignore case -i CLI 0.53 s uvp 0.64 s CLI/uvp=1/1.21 🍊 2137 hits exit=0 .uwvz reused
$ rg -n 'v="(bus_stop|traffic_signals)"' osm/japan-dv-ai
$ uvp osm/japan-dv-ai 'v="(bus_stop|traffic_signals)"' -E
✅ regex -E CLI 0.63 s uvp 0.79 s CLI/uvp=1/1.25 🍊 20620 hits exit=0 .uwvz reused
$ rg -n -F 'k="amenity"' osm/japan-dv-ai | rg -F 'v="parking"'
$ uvp osm/japan-dv-ai 'k="amenity"' 'v="parking"' -limit none
✅ drill-down (2 terms) CLI 0.46 s uvp 1.30 s CLI/uvp=1/2.83 🍊 94843 hits exit=0 .uwvz reused
$ rg -n -F 'k="amenity"' osm/japan-dv-ai | rg -v -F 'v="parking"'
$ uvp osm/japan-dv-ai 'k="amenity"' 'v="parking"' -v -limit none
✅ exclude -v CLI 0.46 s uvp 1.33 s CLI/uvp=1/2.89 🍊 38373 hits exit=0 .uwvz reused
$ uvp osm/japan-dv-ai 'k="amenity"' -grep 'v="parking"' -limit none
✅ drill-down (-grep form) CLI - s uvp 1.39 s CLI/uvp=- 94843 hits exit=0 same result as the two-term form?
$ uvp osm/japan-dv-ai 'k="amenity"' -C 3 'k="name"' -limit none
・ 2nd term within ±3 lines -C CLI - s uvp 1.25 s CLI/uvp=- 14301 hits exit=0 no single CLI command means this (rg -C only prints context)
$ uvp osm/japan-dv-ai -seq 'k="amenity",k="name"' -limit none
・ order -seq CLI - s uvp 1.02 s CLI/uvp=- 37125 hits exit=0 no CLI equivalent (needs an awk state machine)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-dv-ai | LC_ALL=C sort | uniq -c | sort -k1,1nr | head -20
$ uvp osm/japan-dv-ai 'k="amenity"' -uniq 'v="([^"]+)"' -head 20 -limit none
✅ tally -uniq top 20 CLI 0.45 s uvp 0.76 s CLI/uvp=1/1.69 🍊 20 hits exit=0 values and counts compared as a set (sort|uniq -c|sort -rn|head)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-dv-ai | LC_ALL=C sort | uniq -c | sort -k1,1nr | head -20
$ uvp osm/japan-dv-ai 'k="amenity"' -uniq 'v="([^"]+)"' -sort count -head 20 -limit none
✅ tally + -sort count CLI 0.45 s uvp 0.89 s CLI/uvp=1/1.98 🍊 20 hits exit=0 ties may order differently (compared as a set)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-dv-ai | LC_ALL=C sort | uniq -c | LC_ALL=C sort -k2 | head -20
$ uvp osm/japan-dv-ai 'k="amenity"' -uniq 'v="([^"]+)"' -sort value -head 20 -limit none
✅ tally + -sort value CLI 0.44 s uvp 0.87 s CLI/uvp=1/1.98 🍊 20 hits exit=0 top 20 by value (compared in LC_ALL=C order)
$ rg -n -F '東京' osm/japan-dv-ai | head -10
$ uvp osm/japan-dv-ai '東京' -head 10
✅ first 10 -head CLI 0.05 s uvp 0.74 s CLI/uvp=1/14.80 🍊 10 hits exit=0 rg stops after emitting 10 lines (early exit); the full-scan form is the next row (sort then head)
$ rg -n -F '東京' osm/japan-dv-ai | tail -10
$ uvp osm/japan-dv-ai '東京' -tail 10
✅ last 10 -tail CLI 0.34 s uvp 0.83 s CLI/uvp=1/2.44 🍊 10 hits exit=0 rg | tail must scan everything to know the end (no early exit)
$ rg -N -F '東京' osm/japan-dv-ai | LC_ALL=C sort | head -10
$ uvp osm/japan-dv-ai '東京' -sort '^(.*)$' -head 10 --no-line-number
✅ sort then first 10 CLI 0.29 s uvp 0.70 s CLI/uvp=1/2.41 🍊 10 hits exit=0 same as rg | sort | head (LC_ALL=C, key = whole line)
$ rg -n -F '東京' osm/japan-dv-ai > cli-test-results/20260915-170224-uvp4/3G/rg_out.txt
$ uvp osm/japan-dv-ai '東京' -out cli-test-results/20260915-170224-uvp4/3G/uvp_out.txt
✅ -out (text) CLI 0.35 s uvp 0.83 s CLI/uvp=1/2.37 🍊 11274 hits exit=0 the two written files compared
$ rg -n -F '東京' osm/japan-dv-ai | gzip -6 > cli-test-results/20260915-170224-uvp4/3G/rg_out.txt.gz
$ uvp osm/japan-dv-ai '東京' -out cli-test-results/20260915-170224-uvp4/3G/uvp_out.txt.gz
✅ -out (.gz) CLI 0.36 s uvp 0.84 s CLI/uvp=1/2.33 🍊 11274 hits exit=0 decompressed, then contents compared
$ sed -n '/東京/p' osm/japan-dv-ai
$ uvp osm/japan-dv-ai '東京' --no-line-number
✅ lines only ↔ sed -n p CLI 9.61 s uvp 0.75 s CLI/uvp=12.81 🟢 11274 hits exit=0 compared against sed's extraction, body only (no line numbers)
$ uvp osm/japan-dv-ai.uwvz '東京'
✅ search the .uwvz alone CLI - s uvp 0.74 s CLI/uvp=- 11274 hits exit=0 searched from the .uwvz without touching the original
$ uvp osm/japan-dv-ai.uwvz -extract -out cli-test-results/20260915-170224-uvp4/3G/extracted
✅ -extract (byte-identical) CLI - s uvp 3.62 s CLI/uvp=- - hits exit=0 free feature; cmp against the original
$ uvp cat osm/japan-dv-ai
✅ cat (raw text) CLI - s uvp 3.65 s CLI/uvp=- - hits exit=0 cmp against the original file
$ uvp osm/japan-dv-ai NO_SUCH_WORD_ZZ9
✅ no match → exit 1 CLI - s uvp 0.66 s CLI/uvp=- 0 hits exit=1 grep-compatible (0 = found, 1 = not found)
$ rg -n '[0-9]{3}-[0-9]{4}"' osm/japan-dv-ai
$ uvp osm/japan-dv-ai '[0-9]{3}-[0-9]{4}"' -E
✅ regex -E (no literal) CLI 2.46 s uvp 1.08 s CLI/uvp=2.28 🟢 14951 hits exit=0 a pattern with no required literal = no prefilter; compare with the regex row
$ rg -n -F 'k="building"' osm/japan-dv-ai
$ uvp osm/japan-dv-ai 'k="building"'
✅ default limit (exit 2 / all hits) CLI 0.83 s uvp 0.59 s CLI/uvp=1.41 🟢 1000000 hits exit=2 cut off at the 1000000-hit limit (rg found 5747705); exactly the limit, in line order, a subset of rg
$ uvp osm/japan-dv-ai '東京' -limit 1000
✅ -limit 1000 CLI - s uvp 0.32 s CLI/uvp=- 1000 hits exit=2 cut off at the 1000-hit limit (rg found 11274); exactly the limit, in line order, a subset of rg
$ uvp osm/japan-dv-ai 'k="building"' -limit none
✅ -limit none (all hits) CLI 0.83 s uvp 2.80 s CLI/uvp=1/3.37 🍊 5747705 hits exit=0 returns everything and matches rg (5747705 hits)?
== 10G: osm/japan-dv-ac 10255315281 bytes
.uwvz present (1164605577 bytes) → reusing it
$ rg -n -F '東京' osm/japan-dv-ac
$ uvp osm/japan-dv-ac '東京'
✅ search (literal) CLI 11.19 s uvp 1.87 s CLI/uvp=5.98 🟢 11393 hits exit=0 .uwvz reused
$ rg -n -i -F tokyo osm/japan-dv-ac
$ uvp osm/japan-dv-ac tokyo -i
✅ ignore case -i CLI 11.13 s uvp 1.83 s CLI/uvp=6.08 🟢 1286 hits exit=0 .uwvz reused
$ rg -n 'v="(bus_stop|traffic_signals)"' osm/japan-dv-ac
$ uvp osm/japan-dv-ac 'v="(bus_stop|traffic_signals)"' -E
✅ regex -E CLI 11.19 s uvp 2.00 s CLI/uvp=5.59 🟢 167126 hits exit=0 .uwvz reused
$ rg -n -F 'k="amenity"' osm/japan-dv-ac | rg -F 'v="parking"'
$ uvp osm/japan-dv-ac 'k="amenity"' 'v="parking"' -limit none
✅ drill-down (2 terms) CLI 11.14 s uvp 3.75 s CLI/uvp=2.97 🟢 2659 hits exit=0 .uwvz reused
$ rg -n -F 'k="amenity"' osm/japan-dv-ac | rg -v -F 'v="parking"'
$ uvp osm/japan-dv-ac 'k="amenity"' 'v="parking"' -v -limit none
✅ exclude -v CLI 11.03 s uvp 3.96 s CLI/uvp=2.79 🟢 168887 hits exit=0 .uwvz reused
$ uvp osm/japan-dv-ac 'k="amenity"' -grep 'v="parking"' -limit none
✅ drill-down (-grep form) CLI - s uvp 3.75 s CLI/uvp=- 2659 hits exit=0 same result as the two-term form?
$ uvp osm/japan-dv-ac 'k="amenity"' -C 3 'k="name"' -limit none
・ 2nd term within ±3 lines -C CLI - s uvp 3.90 s CLI/uvp=- 66778 hits exit=0 no single CLI command means this (rg -C only prints context)
$ uvp osm/japan-dv-ac -seq 'k="amenity",k="name"' -limit none
・ order -seq CLI - s uvp 2.97 s CLI/uvp=- 118625 hits exit=0 no CLI equivalent (needs an awk state machine)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-dv-ac | LC_ALL=C sort | uniq -c | sort -k1,1nr | head -20
$ uvp osm/japan-dv-ac 'k="amenity"' -uniq 'v="([^"]+)"' -head 20 -limit none
✅ tally -uniq top 20 CLI 11.05 s uvp 2.15 s CLI/uvp=5.14 🟢 20 hits exit=0 values and counts compared as a set (sort|uniq -c|sort -rn|head)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-dv-ac | LC_ALL=C sort | uniq -c | sort -k1,1nr | head -20
$ uvp osm/japan-dv-ac 'k="amenity"' -uniq 'v="([^"]+)"' -sort count -head 20 -limit none
✅ tally + -sort count CLI 11.07 s uvp 2.00 s CLI/uvp=5.54 🟢 20 hits exit=0 ties may order differently (compared as a set)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-dv-ac | LC_ALL=C sort | uniq -c | LC_ALL=C sort -k2 | head -20
$ uvp osm/japan-dv-ac 'k="amenity"' -uniq 'v="([^"]+)"' -sort value -head 20 -limit none
✅ tally + -sort value CLI 11.07 s uvp 2.15 s CLI/uvp=5.15 🟢 20 hits exit=0 top 20 by value (compared in LC_ALL=C order)
$ rg -n -F '東京' osm/japan-dv-ac | head -10
$ uvp osm/japan-dv-ac '東京' -head 10
✅ first 10 -head CLI 0.48 s uvp 1.50 s CLI/uvp=1/3.12 🍊 10 hits exit=0 rg stops after emitting 10 lines (early exit); the full-scan form is the next row (sort then head)
$ rg -n -F '東京' osm/japan-dv-ac | tail -10
$ uvp osm/japan-dv-ac '東京' -tail 10
✅ last 10 -tail CLI 10.61 s uvp 1.95 s CLI/uvp=5.44 🟢 10 hits exit=0 rg | tail must scan everything to know the end (no early exit)
$ rg -N -F '東京' osm/japan-dv-ac | LC_ALL=C sort | head -10
$ uvp osm/japan-dv-ac '東京' -sort '^(.*)$' -head 10 --no-line-number
✅ sort then first 10 CLI 11.98 s uvp 1.99 s CLI/uvp=6.02 🟢 10 hits exit=0 same as rg | sort | head (LC_ALL=C, key = whole line)
$ rg -n -F '東京' osm/japan-dv-ac > cli-test-results/20260915-170224-uvp4/10G/rg_out.txt
$ uvp osm/japan-dv-ac '東京' -out cli-test-results/20260915-170224-uvp4/10G/uvp_out.txt
✅ -out (text) CLI 10.98 s uvp 2.00 s CLI/uvp=5.49 🟢 11393 hits exit=0 the two written files compared
$ rg -n -F '東京' osm/japan-dv-ac | gzip -6 > cli-test-results/20260915-170224-uvp4/10G/rg_out.txt.gz
$ uvp osm/japan-dv-ac '東京' -out cli-test-results/20260915-170224-uvp4/10G/uvp_out.txt.gz
✅ -out (.gz) CLI 11.06 s uvp 1.86 s CLI/uvp=5.95 🟢 11393 hits exit=0 decompressed, then contents compared
$ sed -n '/東京/p' osm/japan-dv-ac
$ uvp osm/japan-dv-ac '東京' --no-line-number
✅ lines only ↔ sed -n p CLI 14.01 s uvp 1.81 s CLI/uvp=7.74 🟢 11393 hits exit=0 compared against sed's extraction, body only (no line numbers)
$ uvp osm/japan-dv-ac.uwvz '東京'
✅ search the .uwvz alone CLI - s uvp 1.55 s CLI/uvp=- 11393 hits exit=0 searched from the .uwvz without touching the original
$ uvp osm/japan-dv-ac.uwvz -extract -out cli-test-results/20260915-170224-uvp4/10G/extracted
✅ -extract (byte-identical) CLI - s uvp 12.93 s CLI/uvp=- - hits exit=0 free feature; cmp against the original
$ uvp cat osm/japan-dv-ac
✅ cat (raw text) CLI - s uvp 38.47 s CLI/uvp=- - hits exit=0 cmp against the original file
$ uvp osm/japan-dv-ac NO_SUCH_WORD_ZZ9
✅ no match → exit 1 CLI - s uvp 1.95 s CLI/uvp=- 0 hits exit=1 grep-compatible (0 = found, 1 = not found)
$ rg -n '[0-9]{3}-[0-9]{4}"' osm/japan-dv-ac
$ uvp osm/japan-dv-ac '[0-9]{3}-[0-9]{4}"' -E
✅ regex -E (no literal) CLI 14.38 s uvp 2.43 s CLI/uvp=5.92 🟢 24988 hits exit=0 a pattern with no required literal = no prefilter; compare with the regex row
$ rg -n -F 'k="building"' osm/japan-dv-ac
$ uvp osm/japan-dv-ac 'k="building"'
✅ default limit (exit 2 / all hits) CLI 11.13 s uvp 1.92 s CLI/uvp=5.80 🟢 861 hits exit=0 limit (?) not reached, all hits returned (same 861 as rg)
$ uvp osm/japan-dv-ac '東京' -limit 1000
✅ -limit 1000 CLI - s uvp 0.63 s CLI/uvp=- 1000 hits exit=2 cut off at the 1000-hit limit (rg found 11393); exactly the limit, in line order, a subset of rg
$ uvp osm/japan-dv-ac 'k="building"' -limit none
✅ -limit none (all hits) CLI 11.13 s uvp 1.47 s CLI/uvp=7.57 🟢 861 hits exit=0 returns everything and matches rg (861 hits)?
== 50G: osm/japan-latest.osm 51254526392 bytes
.uwvz present (5737517251 bytes) → reusing it
$ rg -n -F '東京' osm/japan-latest.osm
$ uvp osm/japan-latest.osm '東京'
✅ search (literal) CLI 55.66 s uvp 7.15 s CLI/uvp=7.78 🟢 94979 hits exit=0 .uwvz reused
$ rg -n -i -F tokyo osm/japan-latest.osm
$ uvp osm/japan-latest.osm tokyo -i
✅ ignore case -i CLI 56.26 s uvp 7.18 s CLI/uvp=7.84 🟢 14355 hits exit=0 .uwvz reused
$ rg -n 'v="(bus_stop|traffic_signals)"' osm/japan-latest.osm
$ uvp osm/japan-latest.osm 'v="(bus_stop|traffic_signals)"' -E
✅ regex -E CLI 55.70 s uvp 7.75 s CLI/uvp=7.19 🟢 602805 hits exit=0 .uwvz reused
$ rg -n -F 'k="amenity"' osm/japan-latest.osm | rg -F 'v="parking"'
$ uvp osm/japan-latest.osm 'k="amenity"' 'v="parking"' -limit none
✅ drill-down (2 terms) CLI 55.33 s uvp 21.48 s CLI/uvp=2.58 🟢 460467 hits exit=0 .uwvz reused
$ rg -n -F 'k="amenity"' osm/japan-latest.osm | rg -v -F 'v="parking"'
$ uvp osm/japan-latest.osm 'k="amenity"' 'v="parking"' -v -limit none
✅ exclude -v CLI 55.35 s uvp 21.10 s CLI/uvp=2.62 🟢 935987 hits exit=0 .uwvz reused
$ uvp osm/japan-latest.osm 'k="amenity"' -grep 'v="parking"' -limit none
✅ drill-down (-grep form) CLI - s uvp 20.34 s CLI/uvp=- 460467 hits exit=0 same result as the two-term form?
$ uvp osm/japan-latest.osm 'k="amenity"' -C 3 'k="name"' -limit none
・ 2nd term within ±3 lines -C CLI - s uvp 19.88 s CLI/uvp=- 544212 hits exit=0 no single CLI command means this (rg -C only prints context)
$ uvp osm/japan-latest.osm -seq 'k="amenity",k="name"' -limit none
・ order -seq CLI - s uvp 13.46 s CLI/uvp=- 846622 hits exit=0 no CLI equivalent (needs an awk state machine)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-latest.osm | LC_ALL=C sort | uniq -c | sort -k1,1nr | head -20
$ uvp osm/japan-latest.osm 'k="amenity"' -uniq 'v="([^"]+)"' -head 20 -limit none
✅ tally -uniq top 20 CLI 55.40 s uvp 8.33 s CLI/uvp=6.65 🟢 20 hits exit=0 values and counts compared as a set (sort|uniq -c|sort -rn|head)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-latest.osm | LC_ALL=C sort | uniq -c | sort -k1,1nr | head -20
$ uvp osm/japan-latest.osm 'k="amenity"' -uniq 'v="([^"]+)"' -sort count -head 20 -limit none
✅ tally + -sort count CLI 55.25 s uvp 8.94 s CLI/uvp=6.18 🟢 20 hits exit=0 ties may order differently (compared as a set)
$ rg -N -o 'k="amenity" v="([^"]+)"' -r '$1' osm/japan-latest.osm | LC_ALL=C sort | uniq -c | LC_ALL=C sort -k2 | head -20
$ uvp osm/japan-latest.osm 'k="amenity"' -uniq 'v="([^"]+)"' -sort value -head 20 -limit none
✅ tally + -sort value CLI 55.28 s uvp 8.42 s CLI/uvp=6.57 🟢 20 hits exit=0 top 20 by value (compared in LC_ALL=C order)
$ rg -n -F '東京' osm/japan-latest.osm | head -10
$ uvp osm/japan-latest.osm '東京' -head 10
✅ first 10 -head CLI 0.16 s uvp 6.74 s CLI/uvp=1/42.12 🍊 10 hits exit=0 rg stops after emitting 10 lines (early exit); the full-scan form is the next row (sort then head)
$ rg -n -F '東京' osm/japan-latest.osm | tail -10
$ uvp osm/japan-latest.osm '東京' -tail 10
✅ last 10 -tail CLI 55.13 s uvp 8.10 s CLI/uvp=6.81 🟢 10 hits exit=0 rg | tail must scan everything to know the end (no early exit)
$ rg -N -F '東京' osm/japan-latest.osm | LC_ALL=C sort | head -10
$ uvp osm/japan-latest.osm '東京' -sort '^(.*)$' -head 10 --no-line-number
✅ sort then first 10 CLI 55.26 s uvp 7.22 s CLI/uvp=7.65 🟢 10 hits exit=0 same as rg | sort | head (LC_ALL=C, key = whole line)
$ rg -n -F '東京' osm/japan-latest.osm > cli-test-results/20260915-170224-uvp4/50G/rg_out.txt
$ uvp osm/japan-latest.osm '東京' -out cli-test-results/20260915-170224-uvp4/50G/uvp_out.txt
✅ -out (text) CLI 55.90 s uvp 7.60 s CLI/uvp=7.36 🟢 94979 hits exit=0 the two written files compared
$ rg -n -F '東京' osm/japan-latest.osm | gzip -6 > cli-test-results/20260915-170224-uvp4/50G/rg_out.txt.gz
$ uvp osm/japan-latest.osm '東京' -out cli-test-results/20260915-170224-uvp4/50G/uvp_out.txt.gz
✅ -out (.gz) CLI 55.95 s uvp 7.28 s CLI/uvp=7.69 🟢 94979 hits exit=0 decompressed, then contents compared
$ sed -n '/東京/p' osm/japan-latest.osm
$ uvp osm/japan-latest.osm '東京' --no-line-number
✅ lines only ↔ sed -n p CLI 105.75 s uvp 12.31 s CLI/uvp=8.59 🟢 94979 hits exit=0 compared against sed's extraction, body only (no line numbers)
$ uvp osm/japan-latest.osm.uwvz '東京'
✅ search the .uwvz alone CLI - s uvp 7.90 s CLI/uvp=- 94979 hits exit=0 searched from the .uwvz without touching the original
$ uvp osm/japan-latest.osm.uwvz -extract -out cli-test-results/20260915-170224-uvp4/50G/extracted
✅ -extract (byte-identical) CLI - s uvp 61.76 s CLI/uvp=- - hits exit=0 free feature; cmp against the original
$ uvp cat osm/japan-latest.osm
✅ cat (raw text) CLI - s uvp 243.12 s CLI/uvp=- - hits exit=0 cmp against the original file
$ uvp osm/japan-latest.osm NO_SUCH_WORD_ZZ9
✅ no match → exit 1 CLI - s uvp 7.03 s CLI/uvp=- 0 hits exit=1 grep-compatible (0 = found, 1 = not found)
$ rg -n '[0-9]{3}-[0-9]{4}"' osm/japan-latest.osm
$ uvp osm/japan-latest.osm '[0-9]{3}-[0-9]{4}"' -E
✅ regex -E (no literal) CLI 67.60 s uvp 12.88 s CLI/uvp=5.25 🟢 224915 hits exit=0 a pattern with no required literal = no prefilter; compare with the regex row
$ rg -n -F 'k="building"' osm/japan-latest.osm
$ uvp osm/japan-latest.osm 'k="building"'
✅ default limit (exit 2 / all hits) CLI 60.35 s uvp 1.24 s CLI/uvp=48.67 🟢 1000000 hits exit=2 cut off at the 1000000-hit limit (rg found 29114490); exactly the limit, in line order, a subset of rg
$ uvp osm/japan-latest.osm '東京' -limit 1000
✅ -limit 1000 CLI - s uvp 0.52 s CLI/uvp=- 1000 hits exit=2 cut off at the 1000-hit limit (rg found 94979); exactly the limit, in line order, a subset of rg
$ uvp osm/japan-latest.osm 'k="building"' -limit none
✅ -limit none (all hits) CLI 60.35 s uvp 27.12 s CLI/uvp=2.23 🟢 29114490 hits exit=0 returns everything and matches rg (29114490 hits)?
== replace 3G: osm/japan-dv-ai
$ uvp osm/japan-dv-ai '東京' -replace TOKYO -out cli-test-results/20260915-170224-uvp4/3G/uvp_rep.txt
$ rg -N -F --passthru '東京' -r TOKYO osm/japan-dv-ai > cli-test-results/20260915-170224-uvp4/3G/rg_rep.txt
✅ replace -replace ↔ rg --passthru CLI 7.06 s uvp 4.57 s CLI/uvp=1.54 🟢 11,361 hits exit=0 whole output compared byte for byte (count = replaced occurrences)
$ LC_ALL=C sed 's/東京/TOKYO/g' osm/japan-dv-ai > cli-test-results/20260915-170224-uvp4/3G/sed_rep.txt
✅ replace -replace ↔ sed 's/A/B/g' CLI 14.94 s uvp 4.57 s CLI/uvp=3.27 🟢 11,361 hits exit=0 whole output compared byte for byte
$ uvp osm/japan-dv-ai 'v="(bus_stop|traffic_signals)"' -E -replace 'v="X_$1"' -out cli-test-results/20260915-170224-uvp4/3G/uvp_repre.txt
$ rg -N --passthru 'v="(bus_stop|traffic_signals)"' -r 'v="X_$1"' osm/japan-dv-ai > cli-test-results/20260915-170224-uvp4/3G/rg_repre.txt
✅ replace -E $1 ↔ rg --passthru CLI 4.97 s uvp 4.17 s CLI/uvp=1.19 🟢 20,620 hits exit=0 whole output compared byte for byte (count = replaced occurrences)
$ LC_ALL=C sed -E 's/v="(bus_stop|traffic_signals)"/v="X_\1"/g' osm/japan-dv-ai > cli-test-results/20260915-170224-uvp4/3G/sed_repre.txt
✅ replace -E $1 ↔ sed -E CLI 24.81 s uvp 4.17 s CLI/uvp=5.95 🟢 20,620 hits exit=0 whole output compared byte for byte
== replace 10G: osm/japan-dv-ac
$ uvp osm/japan-dv-ac '東京' -replace TOKYO -out cli-test-results/20260915-170224-uvp4/10G/uvp_rep.txt
$ rg -N -F --passthru '東京' -r TOKYO osm/japan-dv-ac > cli-test-results/20260915-170224-uvp4/10G/rg_rep.txt
✅ replace -replace ↔ rg --passthru CLI 26.89 s uvp 13.47 s CLI/uvp=2.00 🟢 11,445 hits exit=0 whole output compared byte for byte (count = replaced occurrences)
$ LC_ALL=C sed 's/東京/TOKYO/g' osm/japan-dv-ac > cli-test-results/20260915-170224-uvp4/10G/sed_rep.txt
✅ replace -replace ↔ sed 's/A/B/g' CLI 26.51 s uvp 13.47 s CLI/uvp=1.97 🟢 11,445 hits exit=0 whole output compared byte for byte
$ uvp osm/japan-dv-ac 'v="(bus_stop|traffic_signals)"' -E -replace 'v="X_$1"' -out cli-test-results/20260915-170224-uvp4/10G/uvp_repre.txt
$ rg -N --passthru 'v="(bus_stop|traffic_signals)"' -r 'v="X_$1"' osm/japan-dv-ac > cli-test-results/20260915-170224-uvp4/10G/rg_repre.txt
✅ replace -E $1 ↔ rg --passthru CLI 26.80 s uvp 14.08 s CLI/uvp=1.90 🟢 167,126 hits exit=0 whole output compared byte for byte (count = replaced occurrences)
$ LC_ALL=C sed -E 's/v="(bus_stop|traffic_signals)"/v="X_\1"/g' osm/japan-dv-ac > cli-test-results/20260915-170224-uvp4/10G/sed_repre.txt
✅ replace -E $1 ↔ sed -E CLI 78.36 s uvp 14.08 s CLI/uvp=5.57 🟢 167,126 hits exit=0 whole output compared byte for byte
== replace 50G: osm/japan-latest.osm
$ uvp osm/japan-latest.osm '東京' -replace TOKYO -out cli-test-results/20260915-170224-uvp4/50G/uvp_rep.txt
$ rg -N -F --passthru '東京' -r TOKYO osm/japan-latest.osm > cli-test-results/20260915-170224-uvp4/50G/rg_rep.txt
✅ replace -replace ↔ rg --passthru CLI 146.65 s uvp 69.75 s CLI/uvp=2.10 🟢 95,328 hits exit=0 whole output compared byte for byte (count = replaced occurrences)
$ LC_ALL=C sed 's/東京/TOKYO/g' osm/japan-latest.osm > cli-test-results/20260915-170224-uvp4/50G/sed_rep.txt
✅ replace -replace ↔ sed 's/A/B/g' CLI 187.74 s uvp 69.75 s CLI/uvp=2.69 🟢 95,328 hits exit=0 whole output compared byte for byte
$ uvp osm/japan-latest.osm 'v="(bus_stop|traffic_signals)"' -E -replace 'v="X_$1"' -out cli-test-results/20260915-170224-uvp4/50G/uvp_repre.txt
$ rg -N --passthru 'v="(bus_stop|traffic_signals)"' -r 'v="X_$1"' osm/japan-latest.osm > cli-test-results/20260915-170224-uvp4/50G/rg_repre.txt
✅ replace -E $1 ↔ rg --passthru CLI 132.74 s uvp 69.95 s CLI/uvp=1.90 🟢 602,805 hits exit=0 whole output compared byte for byte (count = replaced occurrences)
$ LC_ALL=C sed -E 's/v="(bus_stop|traffic_signals)"/v="X_\1"/g' osm/japan-latest.osm > cli-test-results/20260915-170224-uvp4/50G/sed_repre.txt
✅ replace -E $1 ↔ sed -E CLI 405.52 s uvp 69.95 s CLI/uvp=5.80 🟢 602,805 hits exit=0 whole output compared byte for byte
================================================================
matched: 81 mismatched: 0 reference only: 6
Report: /Volumes/BIWIN/26work/UwViewData/cli-test-results/20260915-170224-uvp4/summary.md
y4u@MacBook-Air UwViewData % uvp --help
Usage:
uvp <file> <term> [stages…] [options] investigate (flag syntax)
uvp <file> '<term> | sort … | uniq …' investigate (quoted pipe syntax, same meaning)
uvp convert <file> [-out x.uwvz] build the .uwvz
uvp cat <file> [-out …] print the text
uvp <file.uwvz> -extract [txt|gz|zip] restore the original (free)
uvp <file> <term> -replace <text> print the whole text with matches replaced (sed 's/A/B/g'; Edit license)
uvp -open [<file>] launch the app (and open the file if given)
Stages (run in the order written):
<term> / -grep <term> search. Later searches look only within ±N of the previous stage (drill-down)
-i ignore case -E regex -w whole word -v exclude (2nd stage on) -C N context lines
-uniq <key> tally (key = first capture group of a regex)
-sort <key> sort (lines by key / after -uniq by count or value)
-seq a,b,c flows where a→b→c appear in that order
-head N / -tail N first / last N
Options:
-out <file> output file (.gz / .zip are compressed; anything else is text)
-limit N max hits kept per search (default: the app setting, initially 1,000,000;
0 or none = unlimited. Memory grows with the hit count, so write large results with -out)
-replace <text> print the whole text with every match of the term (-i -E -w allowed) replaced.
The file itself is not changed. The text is literal; with -E, $1 etc. insert groups ($$ = $)
--force overwrite an existing output file
--no-line-number omit line numbers
--csv / --json format for tally results
-open at the start: run in the app / at the end: run, then also open in the app
Exit codes: 0=found 1=not found 2=error
Totals for the items with a ratio (the header chart)
Only the items where CLI/uvp= shows a number — a CLI command does the same job and both timings exist — are collected and summed per size. The four replace rows check the same uvp run against both rg and sed, so uvp’s seconds are counted once.
All 21 items (17 search / tally / output + 4 replace; CLI = rg and sed)
| Size | CLI total | uvp total | CLI/uvp |
|---|---|---|---|
| 3 GB | 73.63 s | 25.27 s | 2.9 |
| 10 GB | 343.19 s | 64.19 s | 5.3 |
| 50 GB | 1,833.33 s | 320.54 s | 5.7 |
18 items without the sed rows (CLI = rg only)
| Size | rg total | uvp total | rg/uvp |
|---|---|---|---|
| 3 GB | 24.27 s | 24.52 s | 1.0 |
| 10 GB | 224.31 s | 62.38 s | 3.6 |
| 50 GB | 1,134.32 s | 308.23 s | 3.7 |
The upper row of the header image is the 21 items, the lower row the 18 rg-only items. All uvp figures reuse an existing .uwvz; the one-time .uwvz build (1.1 s at 3 GB, 11.8 s at 10 GB, 59.6 s at 50 GB) is not included.
Items with no rg / uvp ratio
The items where the log shows CLI/uvp=-. There are 8 per size.
| Item | Reason (as noted in the log) |
|---|---|
drill-down, -grep syntax |
Checks that it gives the same result as writing the two terms in a row; the CLI side is timed on the two-term row |
second term within ±3 lines, -C |
No single CLI command with the same meaning (rg -C only prints context) |
sequence, -seq |
No CLI equivalent (would need an awk state machine) |
search the .uwvz alone |
Searches the .uwvz without reading the original; no CLI equivalent |
-extract |
Only cmp against the original; not timed against gzip -dc because the formats differ |
cat |
Only cmp against the original |
| not found → exit 1 | Exit-code check |
-limit 1000 |
uvp-specific hit limit |
The breakdown — 81 matched, 0 mismatched, 6 reference — is the summary line at the end of the log.
Related
- We shipped a uvp command — the same answers as ripgrep, 7× faster at 10 GB. And at 3 GB too, if the cache is cold
- Keep it as .uwvz instead of gzip — 1/9 the size, searchable in 7 seconds, and free to restore
- Five Questions Against 258 GB and 4.5 Billion Lines — Where grep Loses to a GUI
- grep, uniq -c and sed in a GUI — the CLI-to-UwView-Pro Cheat Sheet
- Benchmarks — huge-file tools compared
