For all those who are really interested in the language and have at least a small amount of experience with it already: Nim is ripe for contributing!
While the main team works on bigger things drawing v2 closer, there's still lots of issues of various magnitude and lots of housekeeping bits that can be done while getting yourself acquainted with the project's structure.
Hi, is Nim a good replacement for python? I write a lot of scripts, but I find that python can get cumbersome to maintain. I'm looking for something with stricter typing and a sane packaging/directory system, but also easy/fast to write. If Nim can work with numpy it's a plus
I think out of the bunch of new system programming languages it sure is the most friendly towards users coming from dynamic programming languages.
My nim solutions were actually more concise than my JS code when doing some Advent Of Code puzzles.
And it has nimpy which is a bridge to the Python world though I haven't used it personally.
The only downside I can think of regarding nim is that the standard library is OK-ish I guess but not great. Still needs some work but I think they are aware of it.
yes, definitely for me nim has replaced python as my go to language even for small tasks.
I find it even more productive than python thanks to the typing system. An often neglected highlight that I appreciate a lot is the fact that static typing + procedure overload has the additional benefit that names of functions tend to be very natural (you can reuse the name of function if the type signature is different). That along with UFCS (universal function call syntax), allows to create great apis in no times. Also template are a very easy way to boost productivity and allow to write great DSLs. See also an old but still relevant and very valid introduction of Nim for Python programmers [0]
It can work with numpy and even better there is a performant numpy replacement called arraymancer. Take into account that since you do not need to be afraid of loops, even working with standard sequences is very performant. See also the SciNim getting started book [1]
Well it depends a bit on your usecases. Nim certainly has a stricter type system, and since it is compiled you can be fairly certain that a program you have compiled won't suddenly start to misbehave (e.g. no JavaScript leftpad issue since it won't pull in libraries on runtime). The packaging system is fine for most purposes, and it recently got better with the introduction of lock-files. Once you get comfortable with the standard library and your favorite packages it's also very fast to write, maybe even more so than Python because it has less boilerplate.
As for numpy there are libraries that allow you to work with Python from Nim, and vice-versa. Not entirely sure how well they work for numpy as I haven't tried that myself. But there are also packages which does similar things in Nim. However packages will probably be your biggest hurdle when converting stuff to Nim. The Nim ecosystem is definitely smaller than Pythons, so it's a bit hit-or-miss if it has the packages you need for your specific usecases. You are more than welcome to join the Discord/IRC/Matrix room and ask around though.
For me it was. Python was everything for me - but too slow. It was a pretty quick to convert software to Nim, and I have never looked back. You don't get the whole ecosystem from python, but you get fast code and easy syntax. We even often use Nim instead of small shell scripting.
Nim goes further there by removing the need for tools like Pandas and numpy. You don't have to write code in "Pandas-style" because native Nim is equally fast and sometimes faster.
> Nim goes further there by removing the need for tools like Pandas and numpy. You don't have to write code in "Pandas-style" because native Nim is equally fast and sometimes faster.
Can you clarify this? Is there an equivalent data stack in Nim? Or is it that nim is lighter-weight because, for example, it permits transforming tabular data into language structures natively (i.e., without library support)?
This issue alone means I can't use it at work, and I haven't put much time and thought into Nim as a whole because I could only use it in private toy projects.
I understand it's mostly the AV vendors fault and the devs shouldn't have to worry about a problem they haven't caused, but for me, this grinds real-life adoption of the whole thing to a halt.
It's not clear to me what they could do. In the thread you linked, it's not just the binaries of the various nim executables, but also user generated binaries. And it appears that even signed binaries are getting flagged by some vendors.
I'm guessing the issue is that some malware writers started using nim, and the antivirus vendors then decided to make heuristics that detect nim generated binaries and call it malware.
This is pretty much it, malware was written in Nim, vendors started fingerprinting those binaries but didn't include any/enough non-malware binaries. This means that the fingerprint is more "this program is written in Nim" and less "this is malware written in Nim".
I hope I didn't come off as too aggressive in my original post, I'm not trying to demand anything, or trying to act like I know anything they don't, I most certainly do not. I just love the language so much, I wish I could use it more. The linked issue doesn't show any traction, and other languages used for malware don't have this problem, so I (probably incorrectly) presumed there was a lack of interest to solve this. I apologize if I came off as demanding.
They have to work with and stay on antivirus vendors more aggressively. You can take new releases and submit them to the antivirus companies to get white-listed.
They each have their own processes, but a lot of them allow going to their website and submitting as false-positives, then you can do follow up e-mails.
Probably best to select the top 10 AV companies, to keep the workload down. Per each release, shouldn't be too bad, though likely someone or a group needs to be designated for the task.
We take it very seriously, but there isn't a whole lot we can do unfortunately. Apart from reporting false positives the only venue we could pursue is applying obfuscation practices used by actual viruses. This of course has its own slew of issues.
Yep, I mean, if the virus defence community took the same blanket, lazy approach with C and C++ compilers because viruses can be written in those, too (shock horror!) they'd be shutdown by some big players very fast.
There is trade offs to both sensitivity and insensitivity. Just because you are used to one approach from your favorite programming language does not mean the other way is "wrong".
Want a consistent style? Use that nice --styleCheck:usages option. No problem at all.
Honestly, I think Nim should keep the current approach just to filter out certain people.
What are the arguments for insensitivity? I'm not used to it, and it seems to my rookie eyes that it only opens venues for confusion (forcing to pay closer attention to symbols to mentally "de duplicate" them), inconsistencies within a same/foreign code base that make joining a project/jumping into an external library just slightly harder, and more vim/emacs/tab/spaces jihads.
It actually is something that gives me pause for thought. In my field of coding (finance) ambiguity is terrifying and eventually you get it wrong.
Maybe it's not so bad in practice, but looking from the outside, I think... really? To be specific, case-insensitivity doesn't bother me, but the automatic conversion from camel-case to snake-case... very much so.
You must remember that Nim is a statically typed compiled language with support for distinct types etc. This means that you wouldn't be able to call the wrong procedure if you didn't provide it the right types. Besides, not allowing similarly named but differently cased identifiers causes less ambiguity, not more.
> the automatic conversion from camel-case to snake-case... very much so.
If an identifier clashes, you get a compile time ambiguity error. It means `is_OK` and `isOk` will report ambiguous identifiers and force you to qualify it or rename it.
This also helps further reduce ambiguity indirectly by forcing sensible naming.
My ideal would be to simply not allow having my_variable and myVariable in the same namespace but given that you can do that in essentially every language I sort of prefer that they have to refer to the same thing.
A bit off-topic, but I'd like to see a language à la rust+gc, with pattern matching, traits, and the functional stuff rust got right, but without all the worries of lifetimes, and maybe some easier trait management (function signatures become very ugly imo).
I've heard the claim that lean is going in this direction.
The ergonomics of both are more oriented towards functional programming rather than imperative. Rust is still primarily imperative with great APIs for functional programming.
It has great GC piggybacking on the huge amount of efforts going into dotnet, each release since dotnet core has just gotten faster and faster, not just GC but many parts of the ecosystem, see e.g. https://devblogs.microsoft.com/dotnet/performance-improvemen...
Function signatures are usually not needed due to the type inference algorithm, so while it is statically typed, you often don’t need to write out the types of functions.
Not just seamless, I want it to use rust libraries natively and basically be rust under the hood. Think coffeescript but for rust. The problem with most people's suggestions will be that it deviates too far from rust (giant VMs) and a new language would need a duplicate ecosystem.
I feel like this is an inevitability at this point. Kotlin is basically this, and C#/Java/Typescript are getting there, albeit with some historical baggage that might make it slightly less ergonomic.
I started programming in Nim around 2015 and still write code in it from time to time. At this point I would have a tough time suggesting the language to anyone for much. While you can certainly accomplish almost any programming task with Nim, how much frustration you will encounter fighting the compiler and broken / under-specified language features and semantics might be a turn-off. There are other major warts with Nim, which I'll outline briefly below.
1) Nim's leadership is awful and has always been historically. Those in charge of managing the community are more interested in their personal reputation's and resumes than they are actually contributing working software to the language's ecosystem. Even the BDFL treats Nim like their own personal compiler R&D playground. There are so many broken features in Nim which never see the light of day or never get removed.
2) Continuing to expand on the point above, the Nim compiler has become insanely difficult to work on. As more and more half-baked features have been added to the language, the situation has only worsened. There are many would-be contributors who have come into the community only to leave completely frustrated or appalled by the lack of interest in improving the situation from those leading the community.
3) The development path is driven by leadership and not the community, and leadership has no plan. This was extremely evident last year when Araq and others were asked for a roadmap for Nim and they admitted to not having one, and basically wrote one up in a forum thread. Araq and dom96 and a few others, especially those who have been brave enough to actually deploy Nim code in production (looking at you status.im) tend to dictate what gets worked on or fixed. There is no coordination amongst folks working on initiatives. You can look at the `fusion` project, which was started by Nim's leadership, drew contributors in, and then was abandoned by leadership while maintainers were left on the hook to keep things afloat until they too dumped the project.
Things have gotten so bad, that a hard fork has been in development since last year, and many of the more senior community members or those who have been around for a bit have migrated there: https://github.com/nim-works/nimskull
Nim's community now consists of mostly new members or folks who weren't really active in it before, but maybe used Nim. Of course, the leadership remains in place, but there are only a handful of folks remaining that have been around for more than a year or two. Mostly those who bit the bullet on using Nim in production or built large projects with Nim and are stuck maintaining them.
> many of the more senior community members or those who have been around for a bit have migrated there
It's true that compiler developers like krux, Clyybber, saem migrated to nimskull, I don't disagree.
> but there are only a handful of folks remaining that have been around for more than a year or two
But I totally disagree with your false statements about "senior" community members (those who don't do compiler development) migrating to nimskull - this is not true at all. There are a lot of "senior" people left in the Nim community.
And again, nimskull isn't a "silver bullet" either. It has the flexibility to change major things in the language just because it's a hard fork and almost no one (except people who develop it) uses it for real projects, whereas Nim has an actual userbase that doesn't want their code to be broken, especially considering the fact that nimskull has already removed a lot of the stdlib libraries, especially async (which there is no alternative to in Nimskull).
Most of the "senior" community members who have migrated to NimSkull are those who liked to work on the compiler. The ones who remains using Nim are those who like to write libraries and projects. I've yet to see anyone create a library or project with NimSkull, although that might be because I'm not too exposed to it. I would love to be proven wrong, but for the time being NimSkull looks like a good idea if they can pull it off. But currently it looks like ripping of the band-aid to get to a better place and currently it's still in the burning and stinging phase before things (hopefully) gets better.
I guess your definition of senior and mine vary. I'd consider senior members anyone willing to work on the compiler, as they are the members with the potential to move the language forward and out of the mess it currently finds itself in.
Are there experienced programmers in the Nim community, sure? Are they able to fix Nim with Nim's leadership in place? Doubtful in my mind based on history.
Also, I apologize if it seemed like I was evangelizing Nimskull or suggesting folks switch to it. I agree it is in no shape to serve a replacement for Nim presently, and may not ever be.
Nothing here is untrue, but from my perspective it's overstated. I don't use discord, but I visit the forum daily, follow most of the RFCs, and spend a lot of time coding in Nim (https://github.com/dsrw/enu). I really like Nim, mostly like its community, and think many more people should be using it.
I'm sure fusion could have been handled better, and for 2021 the roadmap was a bit hazy, but I can't think of any other big missteps. Araq, dom, PMunch, and other senior folks are in the forms helping people and answering questions every day, and my interactions with all of them has been very positive. The big post 1.0 feature was arc/orc, and that was very well communicated. Bugs are being fixed, useful new features are being added, and future plans are being discussed in the open.
And Nim itself is great. The "if it compiles, it works" factor is high, yet I almost never feel like the compiler is fighting me. Simple things are simple (I'm teaching it to a group of 12 year olds), it's incredibly flexible, it's fast, and it's suitable for almost any sort of problem. There's nothing else like it, and I expect I'd continue using it for at least a decade even if it switched into maintenance mode tomorrow. I think it will take at least that long for something better to come along.
The compiler works until you stress the type system or try to take advantage of features that are experimental and then stay that way forever. How many iterations of concepts have we been through at this point? What happened to hot code reloading? Why do we have all these legacy GCs?
The point is the compiler has more and more cruft that makes it increasingly more complex and increasingly unreliable. It also makes it so that bugs are harder and harder to fix. Not to mention that it's basically impossible to track what is going on with the language at any given point in time because there is no roadmap or updated status. It also doesn't help that certain folks like to create ten GitHub issues and RFCs for every perceived slight in the language.
Once again, the problem boils down to a complete lack of leadership or interest in running the show - and instead of empowering the community to make that happen, leadership sits back and acts like everything is great and perfect and pats themselves on the back. When someone complains that it is not, leadership finds a way to defend itself and turn sentiment against the person complaining. It happens over and over again and it's why Nim isn't a success story like languages that actually have empowered communities and leaders that are interested in leading and not just collecting book revenues / making .io games, etc... while actively gatekeeping those who would like to get involved in making things better.
> unless you limittest generics, most people wont even encounter this.
I don't think my totally normal usage of generics for type safety could be called "stretching the limits." My usage of generics includes using a parameter [V] for OpenGL meshes to introduce type safety with the vertex type. I'm pretty sure that's perfectly normal?
Recently one of my friends tried compiling my project - pan[1] on Windows, and for whatever reason - it doesn't work! stating that it "cannot instantiate 'V'", whatever that means. I've had this error annoy me many times before, but I have no clue what triggers it or how to reproduce it reliably. Call this a 'stable implementation.'
I'm not sure there is one - but I guess a language where you're not constantly dealing with horrible error message, an under-specified and broken implementation and a shitty stdlib - might be better, even if it doesn't have some of the nicer frills of Nim.
Ye gods and little fishes, is there a new-ish language suitable for application programming, which has a more-or-less feature-complete design and implementation, and where there's no petty drama going on every other day? And which, to boot, lets you get a job done without requiring/strongly suggesting that you become a part of their vibrant community?
A have a sinking feeling such language exists and is called "C".
This matches my experience with Nim community and it's toxic leadership. The amount of petty drama the community leaders get involved with is unbelievable. Spending a few weeks on their Discord, and observing how the leadership interacts with new/opposing ideas is enough to get a clear picture. It doesn't inspire trust.
Contrast this to, say, Zig, their road-maps and community structure is light years ahead.
What the heck man? This doesn't match my experience at all. And I can't find your username on neither matrix nor discord. Did you just found a chance for a sucker punch and took it? I bet Zig is the wonder land.
Readit News
While the main team works on bigger things drawing v2 closer, there's still lots of issues of various magnitude and lots of housekeeping bits that can be done while getting yourself acquainted with the project's structure.
Here's a few links for your consideration:
* https://github.com/nim-lang/Nim/issues?q=is%3Aissue+is%3Aope...
* https://dev.to/xflywind/best-ways-to-make-your-first-contrib...
* https://nim-lang.github.io/Nim/contributing.html
My nim solutions were actually more concise than my JS code when doing some Advent Of Code puzzles.
And it has nimpy which is a bridge to the Python world though I haven't used it personally.
The only downside I can think of regarding nim is that the standard library is OK-ish I guess but not great. Still needs some work but I think they are aware of it.
I find it even more productive than python thanks to the typing system. An often neglected highlight that I appreciate a lot is the fact that static typing + procedure overload has the additional benefit that names of functions tend to be very natural (you can reuse the name of function if the type signature is different). That along with UFCS (universal function call syntax), allows to create great apis in no times. Also template are a very easy way to boost productivity and allow to write great DSLs. See also an old but still relevant and very valid introduction of Nim for Python programmers [0]
It can work with numpy and even better there is a performant numpy replacement called arraymancer. Take into account that since you do not need to be afraid of loops, even working with standard sequences is very performant. See also the SciNim getting started book [1]
[0]: https://blog.zdsmith.com/posts/nim-for-python-programmers.ht... [1]: https://scinim.github.io/getting-started/
As for numpy there are libraries that allow you to work with Python from Nim, and vice-versa. Not entirely sure how well they work for numpy as I haven't tried that myself. But there are also packages which does similar things in Nim. However packages will probably be your biggest hurdle when converting stuff to Nim. The Nim ecosystem is definitely smaller than Pythons, so it's a bit hit-or-miss if it has the packages you need for your specific usecases. You are more than welcome to join the Discord/IRC/Matrix room and ask around though.
Nim goes further there by removing the need for tools like Pandas and numpy. You don't have to write code in "Pandas-style" because native Nim is equally fast and sometimes faster.
Can you clarify this? Is there an equivalent data stack in Nim? Or is it that nim is lighter-weight because, for example, it permits transforming tabular data into language structures natively (i.e., without library support)?
Deleted Comment
This issue alone means I can't use it at work, and I haven't put much time and thought into Nim as a whole because I could only use it in private toy projects.
I understand it's mostly the AV vendors fault and the devs shouldn't have to worry about a problem they haven't caused, but for me, this grinds real-life adoption of the whole thing to a halt.
I'm guessing the issue is that some malware writers started using nim, and the antivirus vendors then decided to make heuristics that detect nim generated binaries and call it malware.
Probably best to select the top 10 AV companies, to keep the workload down. Per each release, shouldn't be too bad, though likely someone or a group needs to be designated for the task.
If not, it's possible to get reasonably priced code signing certificates.
https://news.ycombinator.com/item?id=31238375
This whole question generates a lot of controversy for some reason
There is trade offs to both sensitivity and insensitivity. Just because you are used to one approach from your favorite programming language does not mean the other way is "wrong".
Want a consistent style? Use that nice --styleCheck:usages option. No problem at all.
Honestly, I think Nim should keep the current approach just to filter out certain people.
Maybe it's not so bad in practice, but looking from the outside, I think... really? To be specific, case-insensitivity doesn't bother me, but the automatic conversion from camel-case to snake-case... very much so.
It prevents bugs where a developer mistakes one variable or proc for another having similar names.
For example in another language you might have variables isReady, isready, is_ready and use the wrong one, leading to a bug.
Nim does not allow defining 3 different variables like that.
If an identifier clashes, you get a compile time ambiguity error. It means `is_OK` and `isOk` will report ambiguous identifiers and force you to qualify it or rename it.
This also helps further reduce ambiguity indirectly by forcing sensible naming.
The benefit is to avoid confusing variables or procs named my_var, myVar and myvar.
I've heard the claim that lean is going in this direction.
It has great GC piggybacking on the huge amount of efforts going into dotnet, each release since dotnet core has just gotten faster and faster, not just GC but many parts of the ecosystem, see e.g. https://devblogs.microsoft.com/dotnet/performance-improvemen...
Pattern matching is first class. See e.g. https://docs.microsoft.com/en-us/dotnet/fsharp/language-refe...
I haven’t programmed with traits, but sounds like interfaces which you can have records implement. https://gist.github.com/robertpi/2964793
Function signatures are usually not needed due to the type inference algorithm, so while it is statically typed, you often don’t need to write out the types of functions.
1) Nim's leadership is awful and has always been historically. Those in charge of managing the community are more interested in their personal reputation's and resumes than they are actually contributing working software to the language's ecosystem. Even the BDFL treats Nim like their own personal compiler R&D playground. There are so many broken features in Nim which never see the light of day or never get removed.
2) Continuing to expand on the point above, the Nim compiler has become insanely difficult to work on. As more and more half-baked features have been added to the language, the situation has only worsened. There are many would-be contributors who have come into the community only to leave completely frustrated or appalled by the lack of interest in improving the situation from those leading the community.
3) The development path is driven by leadership and not the community, and leadership has no plan. This was extremely evident last year when Araq and others were asked for a roadmap for Nim and they admitted to not having one, and basically wrote one up in a forum thread. Araq and dom96 and a few others, especially those who have been brave enough to actually deploy Nim code in production (looking at you status.im) tend to dictate what gets worked on or fixed. There is no coordination amongst folks working on initiatives. You can look at the `fusion` project, which was started by Nim's leadership, drew contributors in, and then was abandoned by leadership while maintainers were left on the hook to keep things afloat until they too dumped the project.
Things have gotten so bad, that a hard fork has been in development since last year, and many of the more senior community members or those who have been around for a bit have migrated there: https://github.com/nim-works/nimskull
Nim's community now consists of mostly new members or folks who weren't really active in it before, but maybe used Nim. Of course, the leadership remains in place, but there are only a handful of folks remaining that have been around for more than a year or two. Mostly those who bit the bullet on using Nim in production or built large projects with Nim and are stuck maintaining them.
My advice would be to look elsewhere.
It's true that compiler developers like krux, Clyybber, saem migrated to nimskull, I don't disagree.
> but there are only a handful of folks remaining that have been around for more than a year or two
But I totally disagree with your false statements about "senior" community members (those who don't do compiler development) migrating to nimskull - this is not true at all. There are a lot of "senior" people left in the Nim community.
And again, nimskull isn't a "silver bullet" either. It has the flexibility to change major things in the language just because it's a hard fork and almost no one (except people who develop it) uses it for real projects, whereas Nim has an actual userbase that doesn't want their code to be broken, especially considering the fact that nimskull has already removed a lot of the stdlib libraries, especially async (which there is no alternative to in Nimskull).
Are there experienced programmers in the Nim community, sure? Are they able to fix Nim with Nim's leadership in place? Doubtful in my mind based on history.
Also, I apologize if it seemed like I was evangelizing Nimskull or suggesting folks switch to it. I agree it is in no shape to serve a replacement for Nim presently, and may not ever be.
I'm sure fusion could have been handled better, and for 2021 the roadmap was a bit hazy, but I can't think of any other big missteps. Araq, dom, PMunch, and other senior folks are in the forms helping people and answering questions every day, and my interactions with all of them has been very positive. The big post 1.0 feature was arc/orc, and that was very well communicated. Bugs are being fixed, useful new features are being added, and future plans are being discussed in the open.
And Nim itself is great. The "if it compiles, it works" factor is high, yet I almost never feel like the compiler is fighting me. Simple things are simple (I'm teaching it to a group of 12 year olds), it's incredibly flexible, it's fast, and it's suitable for almost any sort of problem. There's nothing else like it, and I expect I'd continue using it for at least a decade even if it switched into maintenance mode tomorrow. I think it will take at least that long for something better to come along.
The point is the compiler has more and more cruft that makes it increasingly more complex and increasingly unreliable. It also makes it so that bugs are harder and harder to fix. Not to mention that it's basically impossible to track what is going on with the language at any given point in time because there is no roadmap or updated status. It also doesn't help that certain folks like to create ten GitHub issues and RFCs for every perceived slight in the language.
Once again, the problem boils down to a complete lack of leadership or interest in running the show - and instead of empowering the community to make that happen, leadership sits back and acts like everything is great and perfect and pats themselves on the back. When someone complains that it is not, leadership finds a way to defend itself and turn sentiment against the person complaining. It happens over and over again and it's why Nim isn't a success story like languages that actually have empowered communities and leaders that are interested in leading and not just collecting book revenues / making .io games, etc... while actively gatekeeping those who would like to get involved in making things better.
> how much frustration you will encounter fighting the compiler and broken / under-specified language features and semantics might be a turn-off.
unless you limittest generics, most people wont even encounter this.
> My advice would be to look elsewhere.
that would imply that there is a better alternative
I don't think my totally normal usage of generics for type safety could be called "stretching the limits." My usage of generics includes using a parameter [V] for OpenGL meshes to introduce type safety with the vertex type. I'm pretty sure that's perfectly normal?
Recently one of my friends tried compiling my project - pan[1] on Windows, and for whatever reason - it doesn't work! stating that it "cannot instantiate 'V'", whatever that means. I've had this error annoy me many times before, but I have no clue what triggers it or how to reproduce it reliably. Call this a 'stable implementation.'
[1]: https://github.com/liquidev/pan [2]: https://github.com/liquidev/aglet
Deleted Comment
A have a sinking feeling such language exists and is called "C".
Contrast this to, say, Zig, their road-maps and community structure is light years ahead.