Original motivation for Emlop was a faster/more accurate version of genlop -p, and learning
Rust. It has since gained features and maturity to compete on all fronts. This file compares
genlop-0.30.12, qlop-0.97, and emlop-0.7.1. Please report any outdated/incorrect info using
the issue tracker.
Known emerge log parsers:
- Emlop (Rust) is the one you're reading about.
- Genlop (Perl) is the most well known.
- Qlop (C) is pretty fast and part of a larger toolkit.
- Splat (Perl) looks like Genlop's predecessor, dead upstream.
- Pqlop (Python) was an ambitious rewrite, dead upstream.
- Glop (Haskell) was a simple rewrite, dead upstream.
- Golop (Go) is a recent rewrite apparently abandoned quickly.
- Emwa (C) is a recent addition, time will tell.
- Genloppy (Python) aims for drop-in Genlop compatibility, work in progress.
Rust, Perl, C, Python, Haskell, Go... at least Gentoo doesn't suffer from a language monoculture ;)
Emlop is organised into subcommands, whereas {gen,q}lop only use (possibly conflicting) flags. It
tries to merge functions where that makes sense, for example emlop l combines genlop -l, genlop -e, and genlop -t, because there didn't seem to be a point to separate them. Same thing with
genlop -c and genlop -p which are combined into emlop p.
Emlop output aims to be compact, beautiful, flexible, and easy to read/parse. Qlop is similar to genlop, but did make some outputs more compact.
Default qlop duration output depends on length: 45s -> 3′45″ -> 1:23:45. Machine output
applies to dates and durations at the same time.
| genlop | qlop | emlop | |
|---|---|---|---|
| Output density | sparse | compact | compact |
| Optional headers | no | no | yes |
| Aligned output | some | some | all |
| Optional plain tab alignment | no | no | yes |
| Force color output | no | yes | yes |
| Date output formats | - | rfc3339,ts | many |
| Timezone options | utc | - | utc |
| Duration output formats | text | hms,secs,text | many |
| genlop | qlop | emlop | |
|---|---|---|---|
| Display merges/unmerges | yes | yes | yes |
| Distinguish autoclean/manual unmerges | no | yes | no |
| Distinguish syncs per repository | no | no | yes |
| Display unmerge/sync duration | no | yes | yes |
| Display interrupted/failed merges | no | no | no |
| Display currently installed package's USE/CFLAGS/date | yes | no | no |
| Display merge begin time or end time by default | end only | begin | end |
If the log contains a merge end event without a merge start, qlop displays nothing, genlop displays
a buggy time, and emlop displays the time as ?. Qlop also displays nothing when time jumps
backward. As a result, qlop may report fewer total merges than {gen,em}lop.
Qlop sync duration only corresponds to the first repo (typically gentoo). Emlop sync duration
ignores the pre-sync setup time (usually 0 or 1 seconds).
Emlop has a dedicated stats command. {gen,q}lop spread the functionality between multiple and
sometimes incompatible flags.
| genlop | qlop | emlop | |
|---|---|---|---|
| Individual merge count/total/average/prediction | c,t,a | c,t,a | c,t,p |
| Total merge count/total/average | - | c,t,a | c,t,a |
| Total unmerge count/total/average | - | c,t,a | c,t,a |
| Total sync count/total/average/prediction | - | c,t,a | c,t,p |
| Group stats by year/month/week/day | no | no | yes |
Genlop switches case-sensitivity using -s vs -S flag. Emlop doesn't have a flag, but regexp can
be prepended with (?-i) should case-sensitivity ever be needed. Qlop only supports plaintext
whole-word matching.
Genlop and qlop use a single flag for min/max date, so it isn't possible to specify only a max date.
For relative dates, genlop accepts fancy strings like "last month" or "2 weeks ago", qlop is a bit less flexible but less verbose (no "ago" needed), and emlop only accepts a number of days/weeks/etc which can be abbreviated (for example "1 week, 3 days" -> "1w3d").
| genlop | qlop | emlop | |
|---|---|---|---|
| Limit log parsing by date | yes | yes | yes |
| Limit log to number fisrt/last n entries | no | no | yes |
| Limit log to nth emerge operation | no | last only | yes |
| Filter by package categ/name | yes | yes | yes |
| Filter by sync repo | no | no | yes |
| Read filter list from file | no | yes | no |
| Search modes | plain/regex | plain | plain/regex |
| Default search mode | plain | plain | regex |
Genlop uses the mean of the last 10 builds, ignoring the worst/best times. Qlop uses the mean of the last 20 builds. Emlop uses the median of the last 15 builds, with options for other window sizes and other averages (median/mean/weighted). Using a window mitigates against evolving build times, using a median mitigates against exceptional build times. The Emlop defaults have been measured to give significantly better accuracy over a full emerge log.
Qlop can only predict the current merge. Genlop and Emlop can also predict pretended merges (the
output of emerge -p foo). Emlop by default predicts the current full merge list (similar to what
piping emerge -rOp would do).
Genlop has multiple estimation bugs where data get mixed up (different categories, parallel merges,
etc). Genlop -p doesn't take current elapsed emerge time into account. When run as a normal user,
qlop warns about missing /proc permissions, finds bogus current merges, and doesn't give the same
ETA for the ones it finds. The linuxhowtos db is unmaintained and unlikely to contain info for your
CPU and ebuilds.
All tools give pessimistic prediction when packages are merged in parallel, because they assume sequential merging. Even if they detected an ongoing parallel merge, it's not clear how they would estimate the resulting speedup factor.
| genlop | qlop | emlop | |
|---|---|---|---|
| Show ongoing merge ETA | current build | current build | whole list |
Show emerge -p merges ETA |
yes | no | yes |
| Show individual merge ETAs | no | no | yes |
| Show current merge stage | no | no | yes |
| Global ETA format | total time | total time | total time, end date |
| Estimation accuracy | ok | better | best, configurable |
| Query gentoo.linuxhowtos.org for unknown packages | yes | no | no |
Here are timings for some common commands (in milliseconds, 95th centile of 50 runs, see
benches/stdcomp.sh) on a Ryzen 7 4700U with an SSD and the benches/emerge.log file with ~10K
merges.
The commands were selected to be comparable, but Some differences do influence timings. Emlop always show merge time and package version in "log" mode, and looks up portage resume data in "predict" mode. Genlop can't show unmerges of specific package only. Qlop -r still searches the log for unfinished merges when it doesn't find an ongoing merge. Filtering by plaintext isn't noticeably faster than by case-(in)sensitive regexp ({gen,em}lop only).
| genlop | qlop | emlop | |
|---|---|---|---|
genlop -l; qlop -m; emlop l |
701 | 85 | 61 |
genlop -lut; qlop -muUvt; emlop l -smu |
931 | 140 | 109 |
genlop gcc; qlop -m gcc; emlop l -e gcc |
648 | 34 | 19 |
genlop -r --date 2020-10-08; qlop -stl; emlop l -ss -n |
625 | 65 | 12 |
emerge dummybuild&;genlop -c;qlop -r;emlop p |
760 | 77 | 70 |
genlop -p < emerge-p.gcc.out; emlop p < emerge-p.gcc.out |
669 | n/a | 46 |
genlop -p < emerge-p.qt.out; emlop p < emerge-p.qt.out |
3383 | n/a | 48 |
genlop -p < emerge-p.kde.out; emlop p < emerge-p.kde.out |
20063 | n/a | 46 |
Emlop is faster than qlop, which is already comfortably fast (the wall time is often dominated by
the terminal emulator). Genlop is noticably slow for basic tasks, and can be prohibitively slow for
emerge -p ETAs.
| genlop | qlop | emlop | |
|---|---|---|---|
| Shell completion | bash | none | bash/zsh/fish |
| Complete package name | gentoo repo | n/a | any merged |
| Configuration file | no | no | yes |
| Read compressed emerge.log | yes | no | yes |
| Unittests | no | yes | yes |
| Documentation and help | ok | good | good |
| Development activity | minimal | active | active |