Coloured logs by ewels · Pull Request #4573 · nextflow-io/nextflow

ewels · 2023-12-11T23:19:42Z

To do:

Don't use ANSI for the header when -ansi false is set
Respect env var NO_COLOR
Try harder to auto-detect tty / ansi support and switch off automatically
Look into adding colour to the task label (bit in brackets at the end)
More testing, refine colour use
Try to get exact Nextflow green RGB, at least for the header

See: #3976

pditommaso commented on May 29
That's very beautiful, but not sure exist something so cool as Rich for JVM world

pditommaso commented on May 30
Phil. I think this a great opportunity to improve your Groovy skills 😆

I was never going to be able to back down from such a challenge 😆

CleanShot.2023-12-12.at.00.01.19-converted.mp4

Does a few things:

Truncates the start / middle of the process name (instead of the end), so we can see the end. This is usually the interesting bit.
Hides processes with zero tasks if there are more rows to print than available in the terminal (with a message at the end saying how many are hidden)
Drops the (process > ) at < 180 cols width to save horizontal space
Drop the % age completion at < 120 cols to save horizontal space
Has oodles of colour 🌈 🌟

Known limitations:

Code is probably terrible style
Some of it isn't switched off when ANSI is set to false
Likely some nasty ansi codes will go to log files when there isn't a tty attached, should try to auto-detect and disable
Probably doesn't respect NO_COLOR env variables etc.

netlify · 2023-12-11T23:19:47Z

✅ Deploy Preview for nextflow-docs-staging canceled.

Name	Link
🔨 Latest commit	`10049b2`
🔍 Latest deploy log	https://app.netlify.com/sites/nextflow-docs-staging/deploys/65ca587bb940ba0008560ac5

pditommaso · 2023-12-12T06:46:31Z

ewels · 2023-12-12T08:52:41Z

Wondering if we can use the isAnsiSupported() function from jline to automatically disable ANSI log output when running in a non-interactive shell 🤔

https://jline.github.io/jline2/apidocs/reference/jline/Terminal.html#isAnsiSupported()

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Skip processes with zero tasks if we are out of display rows Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

ewels · 2023-12-12T08:58:49Z

Updates since the video above:

Percentage only green when it gets to 100%
Processes with zero tasks hidden if we run out of terminal
Switched ... with … to save a couple of characters width

bentsherman · 2023-12-12T13:22:21Z

Phil, I had an idea about how we could improve the percentage tracking. Until a process receives all the tasks it will execute, the percentage is always overestimating the true percentage. What if we show the percentage as ??? or - at first, and when a process receives the poison pill (which usually happens long before it finishes all of its tasks), the ANSI log can receive that signal and start showing the percentage

bentsherman · 2023-12-12T13:25:24Z

I already implemented this event hook for array jobs and automatic cleanup, so just need to copy some code from there

ewels · 2023-12-13T21:16:37Z

@bentsherman I go back and forth on this a little. I'm not sure which is best, the current output:

[-        ] NFCORE_RNASEQ:RNASEQ:FASTQ_FASTQC_UMITOOLS_TRIMGALORE:TRIMGALORE    [ 75%] 3 of 4 ✔
[-        ] NFCORE_RNASEQ:RNASEQ:ALIGN_STAR:STAR_ALIGN                          [ 33%] 1 of 3 ✔

or what you're proposing:

[-        ] NFCORE_RNASEQ:RNASEQ:FASTQ_FASTQC_UMITOOLS_TRIMGALORE:TRIMGALORE    [ 75%] 3 of 4 ✔
[-        ] NFCORE_RNASEQ:RNASEQ:ALIGN_STAR:STAR_ALIGN                          [ ???] 1 of ? ✔

Even if the of 3 is a lie, at least you can see how many tasks are currently running there. With 1 of ? you have no idea.

I guess you could go halfway and remove the percentages and keep the of n:

[-        ] NFCORE_RNASEQ:RNASEQ:FASTQ_FASTQC_UMITOOLS_TRIMGALORE:TRIMGALORE    [ 75%] 3 of 4 ✔
[-        ] NFCORE_RNASEQ:RNASEQ:ALIGN_STAR:STAR_ALIGN                          [ ???] 1 of 3 ✔

..but I sort of feel that you're still "lying" about the proportion of tasks done. So not sure how much of an improvement it is over the current behaviour.

Am I missing something / wrong about anything here?

bentsherman · 2023-12-13T21:29:55Z

The third option is what I'm proposing, but without the check mark:

[-        ] NFCORE_RNASEQ:RNASEQ:FASTQ_FASTQC_UMITOOLS_TRIMGALORE:TRIMGALORE    [ 75%] 3 of 4 ✔
[-        ] NFCORE_RNASEQ:RNASEQ:ALIGN_STAR:STAR_ALIGN                          [ ???] 1 of 3

The question marks in the percentage and the lack of a check mark tell you that this process hasn't received all of its tasks yet, so the true progress is unknown. I think that's reasonable

ewels · 2023-12-13T21:35:29Z

Ah sorry, the check mark was me manually copying and pasting stuff for the example. That's not there in the current behaviour (see above screenshots).

bentsherman · 2023-12-13T21:49:03Z

I see, well my point is basically the same

ewels · 2023-12-13T21:49:55Z

Yup, agreed. I like it! Feel free to push whatever to this branch to add it 👍🏻

Signed-off-by: Ben Sherman <bentshermann@gmail.com>

bentsherman · 2023-12-14T15:51:42Z

Okay that is done. Another idea I had related to this is to not include failed tasks in the percentage and total count. When a task fails and is retried, the percentage increases slightly (e.g. it goes from 3/4 = 75% to 4/5=80%), giving the illusion that the pipeline made progress when it didn't.

Instead, the number of failed tasks should not influence either the total or completed count, they should just be noted to the side as they already are.

bentsherman · 2023-12-14T16:18:17Z

Here's an example process that runs 4 tasks where 1 task fails and retries:

[70/0de758] process > bar (3) [ 75%] 3 of 4, retries: 1

Then the process shows 100% when the retried task succeeds:

[70/0de758] process > bar (3) [100%] 4 of 4, retries: 1 ✔

If instead the failed task is ignored, it simply reduces the number of total tasks, allowing the process to reach 100% while still noting the number of ignored tasks:

[f2/decd6c] process > bar (4) [ 75%] 2 of 3, ignored: 1

[98/47ff99] process > bar (4) [100%] 3 of 3, ignored: 1 ✔

pditommaso · 2023-12-14T16:26:19Z

Can we keep change in the logic separated from the colouring?

bentsherman · 2023-12-14T16:38:56Z

You mean like a separate PR? Sure

pditommaso · 2023-12-14T16:42:15Z

Thanks

ewels · 2023-12-14T18:59:34Z

@bentsherman are you ok to yank your commits out of this PR again? I'll leave it to you to handle that bit. Shout if you want me to do anything.

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

pditommaso · 2024-02-10T16:31:29Z

I've realised that on older version of org.fusesource.jansi.Ansi embedded with jline.

Keeping both on the classpath sounds a very bad thing, and version 2.4 was only used for the bgRgb and fgRbg methods. Think this should do the same.

modules/nextflow/src/main/groovy/nextflow/trace/AnsiLogObserver.groovy

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

modules/nextflow/src/main/groovy/nextflow/trace/AnsiLogObserver.groovy

pditommaso · 2024-02-10T16:51:32Z

modules/nextflow/src/main/groovy/nextflow/trace/AnsiLogObserver.groovy

+        final labelSpaces = tagMatch ? tagMatch.group(2) : ''
+        final labelNoTag = LBL_REPLACE.matcher(label).replaceFirst("")
+        final labelFinalProcess = labelNoTag.tokenize(':')[-1]
+        final labelNoFinalProcess = labelFinalProcess.length() > 0 ? labelNoTag - ~/$labelFinalProcess$/ : labelNoTag


What labelNoTag - ~/$labelFinalProcess$/ does? just strips labelFinalProcess from labelNoTag?

Yup - I'm basically breaking that string into components so that I can colour them differently. I wanted to highlight the final process name (after the last :) as that's usually the most interesting bit.

I added a chunk of comments a bit further up with an example, and what each variable will end up with.

I think the regex is not necessary, have a look at feb0cc5

ewels · 2024-02-10T18:46:20Z

modules/nextflow/src/main/groovy/nextflow/cli/CmdRun.groovy

+
+            def fmt = ansi()
+            fmt.a("\n")
+            fmt.bg(Color.CYAN).fg(Color.BLACK).bold().a(" N E X T F L O W ").reset()


@pditommaso this isn't quite the same as the bgRgb that was here before. The previous RGB value was the exact Nextflow colour and would appear the same on every terminal - irrespective of what theme is being used. The same for the fgRbg with (0,0,0).

The problem with terminal colours is that "cyan" and "black" are interpreted differently by everyone's setup, depending on what theme they use. For example, in a dark mode termainal, "black" is actually... white 👀 By using the exact rgb values it always looks the same and it ensures that it'll be readable.

This is minor, but I wanted to point out why I did this particular bit the way I did (and why it was different to everything else).

To illustrate this, the output I get now looks like this: (not green and poor text contrast)

Whereas before it looked like this:

Umm, this stuff is little more of a wrapper around an escape code. Considering this is the only place where it's used, is there any chance to find out the code and use it directly?

Ok - Yup, can do 👍🏻

Ok, got it to work after a bit of trial and error in 32e3a16

Now looks exactly as in the second screenshot above.

Not working for me. Either we give up with the background color or we may need to use a different library for Ansi colors. This can be a better choice

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Use raw ANSI codes to get precise RGB codes. Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

ewels · 2024-02-12T14:39:44Z

Ok, I see some difference in the Nextflow header RGB support based depending on which terminal program I use. Mac OSX basically doesn't support the background:

I guess that this is due to Apple Terminal lacking support. I'll try to take a look.

pditommaso · 2024-02-12T15:51:23Z

I suspect they have different escape codes ..

ewels · 2024-02-12T16:03:14Z

I think it's because the Apple Terminal program only supports 256 colours, rather than truecolor (16.7 million). See this blog post, but I've also come across it before in the Python world I remember now.

Basically, Apple Terminal is kind of crap 😅

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

ewels · 2024-02-12T16:57:26Z

Ok, with a lot of trial and error and some help from ChatGPT, I think I found a solution which is close enough to the "true" Nextflow green that no-one will notice the difference, and is hopefully pretty consistent across terminals:

pditommaso · 2024-02-12T17:38:24Z

You went wild! 😆

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

pditommaso · 2024-02-12T17:43:16Z

Let's do it!

ewels added 6 commits December 12, 2023 09:55

Hide 'process >' and %age complete on narrow terminal window sizes

18649eb

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Truncate the start of the process name instead of the end

82ec80b

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

🚀 N E X T F L O W ~ bling header

cecb4af

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Over the top fancy colours for ANSI log

f42c7ee

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Try not to exceed terminal height

eb99f99

Skip processes with zero tasks if we are out of display rows Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Yellow process %ages until 100%, then green

89af89f

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

ewels force-pushed the coloured-logs branch from 89d21ee to 89af89f Compare December 12, 2023 08:55

bentsherman added 3 commits December 13, 2023 23:58

Merge branch 'master' into pr/ewels/4573

e9af857

minor cleanup

89455f2

Signed-off-by: Ben Sherman <bentshermann@gmail.com>

Use Nextflow green in launch header

8ccf02d

Signed-off-by: Ben Sherman <bentshermann@gmail.com>

bentsherman self-requested a review December 14, 2023 14:44

bentsherman force-pushed the coloured-logs branch from d0a84ec to 89455f2 Compare December 14, 2023 16:48

bentsherman mentioned this pull request Dec 14, 2023

Hide percentage progress until process is closed #4588

Closed

Remove dep with jansi 2.4.0 [ci fast]

a39fe9b

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

pditommaso reviewed Feb 10, 2024

View reviewed changes

modules/nextflow/src/main/groovy/nextflow/trace/AnsiLogObserver.groovy Outdated Show resolved Hide resolved

pditommaso added 2 commits February 10, 2024 17:38

add printBanner

7c21657

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

Using pre-compiled pattern to improve efficiency [ci fast]

9090d1e

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

pditommaso reviewed Feb 10, 2024

View reviewed changes

modules/nextflow/src/main/groovy/nextflow/trace/AnsiLogObserver.groovy Show resolved Hide resolved

pditommaso reviewed Feb 10, 2024

View reviewed changes

ewels commented Feb 10, 2024

View reviewed changes

ewels added 2 commits February 10, 2024 20:42

Add a load of comments explaining ANSI log logic

afaf284

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Put back Nextflow green RGB header

32e3a16

Use raw ANSI codes to get precise RGB codes. Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

ewels force-pushed the coloured-logs branch from b20db16 to 32e3a16 Compare February 10, 2024 19:42

ewels and others added 6 commits February 10, 2024 20:44

Fix spelling error spotted by Codespell

36823ac

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Merge branch 'master' into coloured-logs

b8b0c21

Minor change

ee6ee7c

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

Simplify string trimming [ci fast]

feb0cc5

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

Minor changes

40abece

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

Restore Phil code [ci fast]

00c22b1

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

Find ANSI colour codes that work across most (all?) terminals

13e631c

Signed-off-by: Phil Ewels <phil.ewels@seqera.io>

Minor change [ci skip]

10049b2

Signed-off-by: Paolo Di Tommaso <paolo.ditommaso@gmail.com>

pditommaso merged commit 5e2ce9e into nextflow-io:master Feb 12, 2024

ewels deleted the coloured-logs branch February 12, 2024 21:48

ewels mentioned this pull request Aug 22, 2025

Support standard environment variables for ANSI terminal output #6362

Merged

Conversation

ewels commented Dec 11, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

netlify bot commented Dec 11, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

✅ Deploy Preview for nextflow-docs-staging canceled.

Uh oh!

pditommaso commented Dec 12, 2023

Uh oh!

ewels commented Dec 12, 2023

Uh oh!

ewels commented Dec 12, 2023

Uh oh!

bentsherman commented Dec 12, 2023

Uh oh!

bentsherman commented Dec 12, 2023

Uh oh!

ewels commented Dec 13, 2023

Uh oh!

bentsherman commented Dec 13, 2023

Uh oh!

ewels commented Dec 13, 2023

Uh oh!

bentsherman commented Dec 13, 2023

Uh oh!

ewels commented Dec 13, 2023

Uh oh!

bentsherman commented Dec 14, 2023

Uh oh!

bentsherman commented Dec 14, 2023

Uh oh!

pditommaso commented Dec 14, 2023

Uh oh!

bentsherman commented Dec 14, 2023

Uh oh!

pditommaso commented Dec 14, 2023

Uh oh!

ewels commented Dec 14, 2023

Uh oh!

pditommaso commented Feb 10, 2024

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ewels commented Feb 12, 2024

Uh oh!

pditommaso commented Feb 12, 2024

Uh oh!

ewels commented Feb 12, 2024

Uh oh!

ewels commented Feb 12, 2024

Uh oh!

pditommaso commented Feb 12, 2024

Uh oh!

pditommaso commented Feb 12, 2024

Uh oh!

Reviewers

Assignees

Labels

Projects

ewels commented Dec 11, 2023 •

edited

Loading

netlify bot commented Dec 11, 2023 •

edited

Loading