Fork me on GitHub
#clojure-survey2020-01-02
>
Alex Miller (Clojure team)18:01:11

if you magically knew the answer, what would you do with the information?

dominicm20:01:30

Decide where to invest time building tooling?

dominicm21:01:17

Fireplace vs conjure maybe... Although I think users go where the tooling is. So I'm probably working the wrong direction. Wagging the tail of the dog so to speak.

seancorfield18:01:58

Fair question. I don't think it's actionable but I do think it would be interesting to see how much uptake either of those have (REBL, Socket REPL).

Alex Miller (Clojure team)18:01:27

I think I have a sense of REBL from the Patreon/issues/traffic

Alex Miller (Clojure team)18:01:01

I could see socket/nrepl being of interest to @cfleming but not sure who else (but maybe he knows that answer already)

Alex Miller (Clojure team)18:01:01

Q17 What tools do you use to compile/package/deploy/release your Clojure projects? Answers: clj, Leiningen, Boot, Maven, Gradle, Ant, make, custom scripts

Alex Miller (Clojure team)18:01:13

What's missing? What should be removed?

seancorfield18:01:27

@alexmiller Can you share the REBL usage sense you have? (since it sounds like it won't go into the survey).

Alex Miller (Clojure team)18:01:04

it falls off rapidly after that :)

seancorfield18:01:29

That's too bad. I think REBL is awesome and would love to see more folks using it. Of course, I'd also like to see more updates from Cognitect on it too šŸ™‚

Alex Miller (Clojure team)18:01:50

it's small right now and a comparatively narrow slice of attention. lot of chicken and egg there of course. we expected it to be a slow burn.

seancorfield18:01:39

If it were OSS, I'd be eager to contribute šŸ™‚

seancorfield18:01:51

Q17. Given the breadth of packaging tools for clj, is it worth separating some of those out as individual options? (I don't feel strongly about that, just a suggestion). Otherwise, that list sounds good to me.

Alex Miller (Clojure team)18:01:46

like the IDE one, I'm trying to get a big picture feel (doing a separate clj-focused survey at some point might be interesting)

Alex Miller (Clojure team)18:01:02

so trying to get the "families"

Alex Miller (Clojure team)18:01:17

ant was at 0% last year so going to finally remove it :)

4
Alex Miller (Clojure team)18:01:07

this question is on the "Clojure" page too so it omits some cljs possibilities (that's a separate question)

Alex Miller (Clojure team)18:01:03

lots of people put shadow-cljs in the comments for this one, not sure how best to communicate clojure-ness here

Alex Miller (Clojure team)18:01:27

always a bunch of Ansible here too, not sure if that makes sense

Alex Miller (Clojure team)18:01:39

like maybe the question needs different wording here

Alex Miller (Clojure team)18:01:54

maybe I should remove "deploy"

seancorfield18:01:16

For Q15, will 1.10.1 and 1.10.0 be separate options? Or will you just ask about the overall version? (1.10 vs 1.9 etc)

Alex Miller (Clojure team)18:01:42

I changed it to 1.10.x (so grouped). I decided I cared more about 1.10.x vs others than about 1.10.0 vs 1.10.1

4
seancorfield18:01:56

Shadow-cljs is listed in Q22 -- I'm surprised it was "lots" in Q17. Maybe add "(not ClojureScript)` to the question? :)

Alex Miller (Clojure team)18:01:46

you can look in the Other responses - it's easily the most listed

seancorfield18:01:29

Not doubting you, just echoing your surprise, and suggesting a way to avoid it?

ghadi18:01:21

@alexmiller thought about separating compile/package from deploy/release? that's probably why there is a lot of Ansible in the responses

ghadi18:01:21

and I see you said that in the scrollback šŸ™‚

Alex Miller (Clojure team)18:01:33

yeah, maybe it's time to tease this apart a bit more. It could be a matrix question too with tool / use dimensions

Alex Miller (Clojure team)18:01:47

I'm not even sure what the right set of options for "deploy" is but maybe that's a good signal of needing more info

seancorfield19:01:26

deploy has multiple meanings -- we talk about deploying systems to production but we also talk about deploying JAR files to Clojars/Central (and probably other meanings too?)

Alex Miller (Clojure team)19:01:34

I think I'd start with a) what are the right verbs? and b) what are the right nouns? (with the anticipation of maybe changing the whole approach based on the answers)

Alex Miller (Clojure team)19:01:48

yeah, lib vs app are different domains

Alex Miller (Clojure team)19:01:20

for libs, I'm interested in knowing whether you make an artifact, how you make it, and how you put it somewhere for people to use

borkdude19:01:00

I think it would be good to get some statistics on OS usage: Linux, MacOS, Windows, other - maybe split into: local dev environment / production environment

ā¤ļø 4
šŸ‘ 4
borkdude19:01:35

I added: > maybe split into: local dev environment / production environment

borkdude19:01:04

(my impression is that a lot of people develop on MacOS and deploy to linux for example)

Alex Miller (Clojure team)19:01:31

this is worth asking and I will look at the question options for it

Alex Miller (Clojure team)19:01:44

some kind of matrix answer is probably best

seancorfield19:01:39

I get the impression that Windows usage is increasing (for development) as Clojure reaches more people -- based on the number of questions I see about Windows-specific tooling over the last year.

seancorfield19:01:28

Would be nice to have a sense of "primary development O/S" and "primary production O/S" (but I suspect that will be mostly Linux so it may not be as interesting?).

Alex Miller (Clojure team)19:01:17

yeah, I'll split it out although I suspect Linux is the 99% for prod

Alex Miller (Clojure team)19:01:33

Windows dev use is a 2 way thing too - when it's bad, it pushes Windows users away, either forever, or to WSL or Linux VM

Alex Miller (Clojure team)19:01:06

so "there are few Windows users" can be cause or effect

seancorfield19:01:32

It would be nice to finalize CLI on Windows so it could be promoted to the main site, instead of a wiki on an alpha repo šŸ™‚

Alex Miller (Clojure team)19:01:00

yep, it is one of my goals for 2020 :)

borkdude19:01:02

If you want to take any of my ported code (deps.exe stuff), feel free

Alex Miller (Clojure team)19:01:39

yeah, I haven't even looked at any of that yte

seancorfield19:01:42

We should probably take this to #clj-on-windows if we want to dig in further. I've certainly been pretty happy with Scoop as a package manager for Windows but the deps.clj native image has simpler command-line quoting than the PS script so that's also very appealing... And CLI-in-Clojure, compiled to per-platform native code seems nicer than bash + PS + ...

borkdude19:01:02

There is now a single-line PS installer for the executable. The only requirement is Powershell (to run the installer and the checksum code for caching) and java.exe. Works in cmd.exe

borkdude19:01:44

If you have any questions, feel free to ping me in #clj-on-windows or anywhere else

borkdude19:01:53

Just manually downloading the .exe into a dir on the path works too btw.

Filipe Silva11:01:04

on the topic of finalising the windows cli, it would be useful to ask what terminal windows users use

Filipe Silva11:01:49

the windows cli needs to be ran on powershell, but I'm not super confident powershell is what most people use by default

seancorfield17:01:19

Good point! I never used Powershell until the CLI installer required it, and I still do nearly everything from either cmd or WSL instead.

borkdude17:01:53

Luckily there is one option now that supports cmd.exe šŸ™‚

seancorfield17:01:31

Well, the Scoop PS installer adds cmd-clj and cmd-clojure shims that call the PS versions from cmd.exe but I don't know how well quoting works with those...

sogaiu04:01:20

i think it would be interesting to get an idea of wsl use for clojure dev

Alex Miller (Clojure team)04:01:56

yeah, I'm still thinking about how and whether to get that in the survey

Alex Miller (Clojure team)19:01:11

some common answer sets: ā€¢ (for clj git deps): no artifact, n/a, git repo ā€¢ (for Maven artifacts): jar, Leiningen/boot/clj, Leiningen/boot/clj, clojars/s3/deps.co/? ā€¢ (for cljs libs): shadow?, shadow? buck?, npm?

Alex Miller (Clojure team)19:01:02

still not sure I even have the right set of dimensions here, but I think this set of questions could be replaced with something more interesting/useful

Alex Miller (Clojure team)19:01:08

and be more concrete about library vs apps make it able to have more interesting concrete options

Alex Miller (Clojure team)19:01:03

Candidate verbs of interest: ā€¢ manage dependencies ("manage" is weak) ā€¢ compile (I assume this is mostly for apps and in tandem with packaging, so maybe should not be included) ā€¢ run tests ā€¢ package (create an artifact - jar, war, vm package, cloud template ?) ā€¢ release library (make a version of a library available for use) ā€¢ deploy application

cfleming22:01:30

Retrieve dependencies?

Alex Miller (Clojure team)19:01:24

stepping away for a few, feel free to comment on that

pez21:01:53

I think it would be interesting to see Clojure and ClojureScript as separate server side languages.

seancorfield21:01:43

The survey is already partitioned into Clojure-specific, ClojureScript-specific, and common sections...

Alex Miller (Clojure team)21:01:09

he means in the server-side language used with cljs question

Alex Miller (Clojure team)21:01:48

I can break those out. I have no idea what that breakdown would be

seancorfield21:01:11

Q21 gives a partial sense of that, separating Browser from Node.js?

pez21:01:42

I think just having Clojure and ClojureScript as separate options on Q20 would do it.

seancorfield21:01:17

I guess Q21 includes some small group of folks using cljs only on the backend, no frontend...?

pez21:01:44

Also, I am sometimes using Nashorn as my CLJS REPL, so maybe that should be a an option for whatever Q asked about that.

cfleming21:01:18

I may have missed this, but is there a draft of this weā€™re looking at while discussing?

pez21:01:38

We are looking at the 2019 survey. Posted in the start of this channel.

cfleming21:01:47

Ok, thanks.

cfleming21:01:25

Is Q11 useful information for anyone? It seems like that might inform the development of Clojure itself, but itā€™s mostly asking about things which are done now.

cfleming21:01:24

Re: editor choice, itā€™s tricky. In general, I tend to think that Emacs vs non-Emacs is the main interesting question, but Iā€™m not sure the same is true of e.g. Atom. Thinking about it, I tend to have a view of Emacs as a homogeneous environment which is clearly not true despite CIDERā€™s dominance.

eggsyntax22:01:19

I think it's fairly true as far as Clojure/Script dev in emacs goes -- I've only run across a small handful of people who don't use CIDER, and that tends to come hand-in-hand with several other packages. I'm (idly) curious what non-homogeneity you were thinking of.

cfleming22:01:53

Emacs is pretty much the textbook definition of a non-homogeneous environment - every Emacs installation on the planet is basically a special snowflake. But specifically for Clojure, thereā€™s CIDER, clojure-mode without CIDER, inf-clojure, enspiral and probably more Iā€™m not aware of. Someone mentioned monroe, which Iā€™ve seen around but have no idea if anyone beyond the author is using.

eggsyntax22:01:17

Oh, totally agreed that emacs in general is all over the place. But my impression as an emacs user has been that the vast majority of people are using CIDER at this point (98%+, maybe?).

cfleming22:01:13

Thatā€™s my impression too, but I donā€™t really know. Iā€™d be interested to know how many people use clj-refactor, too.

šŸ‘ 8
seancorfield22:01:36

I'd love to know of the Atom users how many are still trying to use ProtoREPL vs how many are now using Chlorine -- but it's purely curiosity, rather than a useful metric. And of course, with the demise of ProtoREPL, there may well have been people abandoning Atom altogether...

cfleming21:01:55

I donā€™t think that clojure-kit warrants a separate question, especially if thereā€™s an Other option which can be filled in manually. I follow their repo so I see all their issues, and thereā€™s very little activity.

Alex Miller (Clojure team)22:01:10

Thatā€™s what I decided too

Alex Miller (Clojure team)22:01:28

There were 3 people that listed it in other last year

cfleming21:01:16

It would be nice if it were possible to have a series of checkboxes or a radio button associated with each option, so selecting Atom would allow you to choose between Chlorine, Proto-REPL or whatever.

dpsutton22:01:10

that's a great idea. i hope the survey allows it

dpsutton22:01:55

i'm interested to see how many use emacs but not CIDER. I know there are a few

Alex Miller (Clojure team)22:01:05

Neither of those are worth breaking out imo

ghadi21:01:33

which one is Q11?

cfleming22:01:22

Which aspects of Clojure have been important to youā€¦

Alex Miller (Clojure team)22:01:19

Actually, I still find this one useful, particularly when compared longitudinally over many years

eggsyntax22:01:54

It'd be interesting to have a free-response question about libraries -- something like "What non-core libraries do you use most?" or "What third-party libraries are most critical for you as a Clojure/Script developer?". I think the answers would be useful for Clojurists Together, and would be useful to the community in helping less community-involved folks keep up with the critical ecosystem as new parts of it emerge. That said, I realize that free-text questions are a pain in the ass to deal with šŸ™‚

cfleming22:01:29

Iā€™d definitely be interested in this too. It would be really useful to have a list of libraries by popularity, but I have no idea how to collect that or how to measure popularity.

eggsyntax22:01:06

I guess number of stars on github is a proxy, but a pretty rough one. It seems like it especially wouldn't capture libraries gradually falling out into disuse (I know I, at least, rarely remove stars...).

dpsutton22:01:23

clojar downloads?

ā˜ļø 4
eggsyntax22:01:18

That definitely seems better than github stars, but also doesn't seems like it captures decreased use over time very well.

Alex Miller (Clojure team)22:01:12

Unless people actually list all of their libraries (and dependent libraries?) it seems unlikely that this would yield meaningful data

Alex Miller (Clojure team)23:01:11

which is to say, I completely agree that if I could wave a wand and know the answer to this, I'd be interested. I've spent time a couple years trying to design a question to do so, and gave up. Either there is public stuff that could be automatically analyzed (pour one out for clj-info RIP) better than a survey OR it's manual which is either too tedious OR it's manual but wildly incomplete.

Alex Miller (Clojure team)23:01:14

you also have to take into account transitive dependencies and lib frequency to really get good information

Alex Miller (Clojure team)23:01:25

the automated approach is the best answer to that imo

cfleming23:01:32

The best I can think of would be to find all project.clj and deps.edn files on Github and parse them. I think Github provides data dumps which might make that tractable, but I donā€™t know enough about it.

Alex Miller (Clojure team)23:01:49

very doable - clj-info used to do this

Alex Miller (Clojure team)23:01:10

sorry cross-clj ? whatever it was calleed

cfleming23:01:09

Oh yeah, cross-clj. That was wider in scope though, it actually parsed all the program source, and only did it for a specific list of projects.

Alex Miller (Clojure team)23:01:47

I think Eric Normand did this analysis too

Alex Miller (Clojure team)23:01:37

I think the danger is in over-reporting direct deps and under-reporting transitive deps, but sounds like a fun project :)

eggsyntax23:01:54

That all makes sense to me, and clearly youā€™ve spent plenty of time thinking about it!

dpsutton22:01:01

@cfleming mentioned a good point. "how many people use clj-refactor". Maybe make this a bit more broad and check static analysis tools? clojure-lsp, clj-refactor, joker, clj-kondo, cursive

cfleming22:01:43

Yes, Iā€™d be very interested in knowing this.

cfleming22:01:06

Although refactoring tools are kind of orthogonal to static analysis.

dpsutton22:01:53

Yeah a multi select. I think all of the refactoring tools require static analysis of some sort though

dpsutton22:01:13

But a general lint/static category

Alex Miller (Clojure team)13:01:16

Iā€™m not planning on expanding the survey to include stuff like this, although I think a lib and dev tool survey would possibly be interesting

cfleming22:01:17

In Q23 Iā€™d like to know how many people are using figwheel main and how many are using figwheel classic.

āž• 8
cfleming22:01:45

Also, thereā€™s only an option for ā€œClojureScriptā€™s browser REPLā€, but it also has REPLs for node.js, nashorn etc. Thereā€™s probably a whole spectrum of potential information to be teased out here, but I donā€™t know how to ask the right questions for ā€œI use the node.js REPL for server side code but the browser REPL is too hard to set upā€.

cfleming22:01:23

Q24 should include Shadow as an option.

šŸ‘ 4
Alex Miller (Clojure team)23:01:00

@cfleming on Q23, I'm wondering if asking this question in a different way would help make it more useful. Something like "What tools do you use for interactive ClojureScript development?" with multi select

cfleming23:01:58

Yes, that sounds better. I think there are too many dimensions to get all the information, but that would be a good start.

Alex Miller (Clojure team)23:01:22

yeah, "REPL" seems very limiting given all the things that are available

Alex Miller (Clojure team)23:01:43

like is Ghostwheel another option too?

cfleming23:01:18

Right, lots of people donā€™t use REPLs at all and just use hot reloading, and that might be distorting the ā€œI donā€™t use a REPLā€ answer.

cfleming23:01:01

I actually donā€™t know the answer to that - Iā€™m aware that Ghostwheel is a thing that exists, but thatā€™s about it.

Alex Miller (Clojure team)23:01:47

I guess ghostwheel is an additional thing that might be used in tandem with more than one answer

dpsutton23:01:51

Ghostwheel replaces defn so specs are inline with the function definition I think. Irrelevant here