I can't speak for parent but I'm currently writing a one-man SaaS using common lisp on the backend and vanilla js on the frontend, and it's been great so far. SBCL's compiler produces very performant code and I find I can be very productive in lisp. The library ecosystem isn't quite as rich as, say, python's but it's still good.
My chief complaint is that if you stray too far off the "happy path" of well-known libraries the quality (or even presence of) the documentation tends to decrease dramatically.
Oh, cool. I’m thinking about something similar with SBCL plus vanilla JS. I’m not worried about libraries and these days try to be a library/framework free as possible. The web dev basics of an application server, access to an RDBMS, and possibly help spitting out HTML/CSS/JS are all I need. And it seems like Common Lisp checks those boxes. I’ll give quicklisp props too as everything I’ve tried to install ran on one go. That’s unfortunately not always my experience elsewhere.
Out of curiosity, have you done Lisp before? I’m curious if productivity gains would be lost in me ramping up.
I've dabbled in lisp on and off for a number of years but have never written it professionally.
Back in the early aughts I learned a much more orthodox way of programming with C++ in first year university and fiddled with a bunch of scripting languages making game mods and stuff in my teens and early 20s. C# is probably what I have written the greatest number of LOC in, so I'm very much an ALGOL normie.
Lisp isn't as weird as people like to pretend it is. Once I wrapped my head around the syntax (which takes maybe an hour) the biggest things I wrestled with were
- the type system (variables don't have types unlike C# and friends, they're just named registers where you can shove anything)
- CLOS (in my opinion it's kind of a leaky abstraction, not as good at encapsulating/hiding the underlying complexity as, say, C# or Java, but CLOS definitely has its strengths)
- the language-adjacent concept and nomenclature like Systems. What Lisp calls Packages are pretty close to what most other languages would call namespaces, and what lisp calls Systems are what other languages call Packages, or maybe libraries.
My experience has been that it's a very expressive language and easy to write once you get into the flow of things.
> (variables don't have types unlike C# and friends, they're just named registers where you can shove anything)
This depends on the implementation one uses: SBCL does type checking:
* (defvar *a-number* 0)
*A-NUMBER*
* (declaim (type (integer 0 100) *a-number*))
(*A-NUMBER*)
* (setf *a-number* 10)
10
* (setf *a-number* "ten")
; in: SETF *A-NUMBER*
; (SETF *A-NUMBER* "ten")
; ==>
; (THE (MOD 101) "ten")
;
; caught WARNING:
; Constant "ten" conflicts with its asserted type (MOD 101).
; See also:
; The SBCL Manual, Node "Handling of Types"
;
; compilation unit finished
; caught 1 WARNING condition
SBCL does that a compile time, too.
> - the language-adjacent concept and nomenclature like Systems. What Lisp calls Packages are pretty close to what most other languages would call namespaces, and what lisp calls Systems are what other languages call Packages, or maybe libraries.
Lisp is old, these names are from end 70s ("package") and ~1981 ("system") - at that time they were used in the MIT Lisp Machine system software.
No, the differences is that in Common Lisp / SBCL type declarations are optional (note though: there is also type inference in SBCL), type declaration are more verbose, type declarations are sometimes a separate expression and a variable by default has the type T, unless the compiler can infer another type.
(declaim (type (integer 0 100) *a-number*))
Above does not declare the type of a value, it declares the type of a variable to be an integer in the range of 0 to 100.
In SBCL we can also query for the type of that variable:
Another example: a function with two local variables. One variable is declared to be of type INTEGER and the other of type STRING. We then try to set one variable to the value of the other:
* (defun foo (a b)
(declare (type integer a)
(type string b))
(setf a b))
; in: DEFUN FOO
; (SETF A B)
; ==>
; (THE INTEGER B)
;
; caught WARNING:
; Derived type of COMMON-LISP-USER::B is
; (VALUES STRING &OPTIONAL),
; conflicting with its asserted type
; INTEGER.
; See also:
; The SBCL Manual, Node "Handling of Types"
;
What you see above is that the SBCL compiler a compile time detects a type error and warns. There are no values involved.
My chief complaint is that if you stray too far off the "happy path" of well-known libraries the quality (or even presence of) the documentation tends to decrease dramatically.