18 files changed, 2366 insertions, 0 deletions

diff --git a/users/sterni/nix/char/all-chars.bin b/users/sterni/nix/char/all-chars.bin
new file mode 100644
index 0000000000..017b909e8e
--- /dev/null
+++ b/users/sterni/nix/char/all-chars.bin
@@ -0,0 +1,2 @@
+
	
+
 !"#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{|}~��������������������������������������������������������������������������������������������������������������������������������
\ No newline at end of file
diff --git a/users/sterni/nix/char/default.nix b/users/sterni/nix/char/default.nix
new file mode 100644
index 0000000000..9c6ce2fb25
--- /dev/null
+++ b/users/sterni/nix/char/default.nix
@@ -0,0 +1,99 @@
+{ depot, lib, pkgs, ... }:
+
+let
+
+  inherit (depot.users.sterni.nix.flow)
+    cond
+    ;
+
+  inherit (depot.nix)
+    yants
+    ;
+
+  inherit (depot.users.sterni.nix)
+    string
+    ;
+
+  # A char is the atomic element of a nix string
+  # which is essentially an array of arbitrary bytes
+  # as long as they are not a NUL byte.
+  #
+  # A char is neither a byte nor a unicode codepoint!
+  char = yants.restrict "char" (s: builtins.stringLength s == 1) yants.string;
+
+  # integer representation of char
+  charval = yants.restrict "charval" (i: i >= 1 && i < 256) yants.int;
+
+  allChars = builtins.readFile ./all-chars.bin;
+
+  # Originally I searched a list for this, but came to the
+  # conclusion that this can never be fast enough in Nix.
+  # We therefore use a solution similar to infinisil's.
+  ordMap = builtins.listToAttrs
+    (lib.imap1 (i: v: { name = v; value = i; })
+      (string.toChars allChars));
+
+  # Note on performance:
+  # chr and ord have been benchmarked using the following cases:
+  #
+  #  builtins.map ord (lib.stringToCharacters allChars)
+  #  builtins.map chr (builtins.genList (int.add 1) 255
+  #
+  # The findings are as follows:
+  # 1. Searching through either strings using recursion is
+  #    unbearably slow in Nix, leading to evaluation times
+  #    of up to 3s for the following very small test case.
+  #    This is why we use the trusty attribute set for ord.
+  # 2. String indexing is much faster than list indexing which
+  #    is why we use the former for chr.
+  ord = c: ordMap."${c}";
+
+  chr = i: string.charAt (i - 1) allChars;
+
+  asciiAlpha = c:
+    let
+      v = ord c;
+    in
+    (v >= 65 && v <= 90)
+    || (v >= 97 && v <= 122);
+
+  asciiNum = c:
+    let
+      v = ord c;
+    in
+    v >= 48 && v <= 57;
+
+  asciiAlphaNum = c: asciiAlpha c || asciiNum c;
+
+in
+{
+  inherit
+    allChars
+    char
+    charval
+    ord
+    chr
+    asciiAlpha
+    asciiNum
+    asciiAlphaNum
+    ;
+
+  # originally I generated a nix file containing a list of
+  # characters, but infinisil uses a better way which I adapt
+  # which is using builtins.readFile instead of import.
+  __generateAllChars = pkgs.runCommandCC "generate-all-chars"
+    {
+      source = ''
+        #include <stdio.h>
+
+        int main(void) {
+          for(int i = 1; i <= 0xff; i++) {
+            putchar(i);
+          }
+        }
+      '';
+      passAsFile = [ "source" ];
+    } ''
+    $CC -o "$out" -x c "$sourcePath"
+  '';
+}
diff --git a/users/sterni/nix/char/tests/default.nix b/users/sterni/nix/char/tests/default.nix
new file mode 100644
index 0000000000..313df47451
--- /dev/null
+++ b/users/sterni/nix/char/tests/default.nix
@@ -0,0 +1,31 @@
+{ depot, ... }:
+
+let
+  inherit (depot.nix.runTestsuite)
+    it
+    assertEq
+    runTestsuite
+    ;
+
+  inherit (depot.users.sterni.nix)
+    char
+    string
+    int
+    fun
+    ;
+
+  charList = string.toChars char.allChars;
+
+  testAllCharConversion = it "tests conversion of all chars" [
+    (assertEq "char.chr converts to char.allChars"
+      (builtins.genList (fun.rl char.chr (int.add 1)) 255)
+      charList)
+    (assertEq "char.ord converts from char.allChars"
+      (builtins.genList (int.add 1) 255)
+      (builtins.map char.ord charList))
+  ];
+
+in
+runTestsuite "char" [
+  testAllCharConversion
+]
diff --git a/users/sterni/nix/flow/default.nix b/users/sterni/nix/flow/default.nix
new file mode 100644
index 0000000000..4bef0abb91
--- /dev/null
+++ b/users/sterni/nix/flow/default.nix
@@ -0,0 +1,83 @@
+{ depot, ... }:
+
+let
+
+  inherit (depot.nix)
+    yants
+    ;
+
+  inherit (depot.users.sterni.nix)
+    fun
+    ;
+
+  # we must avoid evaluating any of the sublists
+  # as they may contain conditions that throw
+  condition = yants.restrict "condition"
+    (ls: builtins.length ls == 2)
+    (yants.list yants.any);
+
+  /* Like the common lisp macro: takes a list
+     of two elemented lists whose first element
+     is a boolean. The second element of the
+     first list that has true as its first
+     element is returned.
+
+     Type: [ [ bool a ] ] -> a
+
+     Example:
+
+     cond [
+       [ (builtins.isString true) 12 ]
+       [ (3 == 2) 13 ]
+       [ true 42 ]
+     ]
+
+     => 42
+   */
+  cond = conds: switch true conds;
+
+  /* Generic pattern match-ish construct for nix.
+     Takes a bunch of lists which are of length
+     two and checks the first element for either
+     a predicate or a value. The second value of
+     the first list which either has a value equal
+     to or a function that evaluates to true for
+     the given value.
+
+     Type: a -> [ [ (function | a) b ] ] -> b
+
+     Example:
+
+     switch "foo" [
+       [ "smol" "SMOL!!!" ]
+       [ (x: builtins.stringLength x <= 3) "smol-ish" ]
+       [ (fun.const true) "not smol" ]
+      ]
+
+      => "smol-ish"
+  */
+  switch = x: conds:
+    if builtins.length conds == 0
+    then builtins.throw "exhausted all conditions"
+    else
+      let
+        c = condition (builtins.head conds);
+        s = builtins.head c;
+        b =
+          if builtins.isFunction s
+          then s x
+          else x == s;
+      in
+      if b
+      then builtins.elemAt c 1
+      else switch x (builtins.tail conds);
+
+
+
+in
+{
+  inherit
+    cond
+    switch
+    ;
+}
diff --git a/users/sterni/nix/flow/tests/default.nix b/users/sterni/nix/flow/tests/default.nix
new file mode 100644
index 0000000000..9f974a61c7
--- /dev/null
+++ b/users/sterni/nix/flow/tests/default.nix
@@ -0,0 +1,39 @@
+{ depot, ... }:
+
+let
+
+  inherit (depot.nix.runTestsuite)
+    runTestsuite
+    it
+    assertEq
+    assertThrows
+    ;
+
+  inherit (depot.users.sterni.nix.flow)
+    cond
+    match
+    ;
+
+  dontEval = builtins.throw "this should not get evaluated";
+
+  testCond = it "tests cond" [
+    (assertThrows "malformed cond list"
+      (cond [ [ true 1 2 ] [ false 1 ] ]))
+    (assertEq "last is true" "last"
+      (cond [
+        [ false dontEval ]
+        [ false dontEval ]
+        [ true "last" ]
+      ]))
+    (assertEq "first is true" 1
+      (cond [
+        [ true 1 ]
+        [ true dontEval ]
+        [ true dontEval ]
+      ]))
+  ];
+
+in
+runTestsuite "nix.flow" [
+  testCond
+]
diff --git a/users/sterni/nix/fun/default.nix b/users/sterni/nix/fun/default.nix
new file mode 100644
index 0000000000..bb10f9e6c1
--- /dev/null
+++ b/users/sterni/nix/fun/default.nix
@@ -0,0 +1,257 @@
+{ depot, lib, ... }:
+
+let
+
+  inherit (lib)
+    id
+    ;
+
+  # Simple function composition,
+  # application is right to left.
+  rl = f1: f2:
+    (x: f1 (f2 x));
+
+  # Compose a list of functions,
+  # application is right to left.
+  rls = fs:
+    builtins.foldl' (fOut: f: lr f fOut) id fs;
+
+  # Simple function composition,
+  # application is left to right.
+  lr = f1: f2:
+    (x: f2 (f1 x));
+
+  # Compose a list of functions,
+  # application is left to right
+  lrs = x: fs:
+    builtins.foldl' (v: f: f v) x fs;
+
+  # Warning: cursed function
+  #
+  # Check if a function has an attribute
+  # set pattern with an ellipsis as its argument.
+  #
+  # s/o to puck for discovering that you could use
+  # builtins.toXML to introspect functions more than
+  # you should be able to in Nix.
+  hasEllipsis = f:
+    builtins.isFunction f &&
+    builtins.match ".*<attrspat ellipsis=\"1\">.*"
+      (builtins.toXML f) != null;
+
+  /* Return the number of arguments the given function accepts or 0 if the value
+     is not a function.
+
+     Example:
+
+       argCount argCount
+       => 1
+
+       argCount builtins.add
+       => 2
+
+       argCount pkgs.stdenv.mkDerivation
+       => 1
+  */
+  argCount = f:
+    let
+      # N.B. since we are only interested if the result of calling is a function
+      # as opposed to a normal value or evaluation failure, we never need to
+      # check success, as value will be false (i.e. not a function) in the
+      # failure case.
+      called = builtins.tryEval (
+        f (builtins.throw "You should never see this error message")
+      );
+    in
+    if !(builtins.isFunction f || builtins.isFunction (f.__functor or null))
+    then 0
+    else 1 + argCount called.value;
+
+  /* Call a given function with a given list of arguments.
+
+     Example:
+
+       apply builtins.sub [ 20 10 ]
+       => 10
+  */
+  apply = f: args:
+    builtins.foldl' (f: x: f x) f args;
+
+  # TODO(sterni): think of a better name for unapply
+  /* Collect n arguments into a list and pass them to the given function.
+     Allows calling a function that expects a list by feeding it the list
+     elements individually as function arguments - the limitation is
+     that the list must be of constant length.
+
+     This is mainly useful for functions that wrap other, arbitrary functions
+     in conjunction with argCount and apply, since lists of arguments are
+     easier to deal with usually.
+
+     Example:
+
+       (unapply 3 lib.id) 1 2 3
+       => [ 1 2 3 ]
+
+       (unapply 5 lib.reverse) 1 2 null 4 5
+       => [ 5 4 null 2 1 ]
+
+       # unapply and apply compose the identity relation together
+
+       unapply (argCount f) (apply f)
+       # is equivalent to f (if the function has a constant number of arguments)
+
+       (unapply 2 (apply builtins.sub)) 20 10
+       => 10
+  */
+  unapply =
+    let
+      unapply' = acc: n: f: x:
+        if n == 1
+        then f (acc ++ [ x ])
+        else unapply' (acc ++ [ x ]) (n - 1) f;
+    in
+    unapply' [ ];
+
+  /* Optimize a tail recursive Nix function by intercepting the recursive
+     function application and expressing it in terms of builtins.genericClosure
+     instead. The main benefit of this optimization is that even a naively
+     written recursive algorithm won't overflow the stack.
+
+     For this to work the following things prerequisites are necessary:
+
+     - The passed function needs to be a fix point for its self reference,
+       i. e. the argument to tailCallOpt needs to be of the form
+       `self: # function body that uses self to call itself`.
+       This is because tailCallOpt needs to manipulate the call to self
+       which otherwise wouldn't be possible due to Nix's lexical scoping.
+
+     - The passed function may only call itself as a tail call, all other
+       forms of recursions will fail evaluation.
+
+     This function was mainly written to prove that builtins.genericClosure
+     can be used to express any (tail) recursive algorithm. It can be used
+     to avoid stack overflows for deeply recursive, but naively written
+     functions (in the context of Nix this mainly means using recursion
+     instead of (ab)using more performant and less limited builtins).
+     A better alternative to using this function is probably translating
+     the algorithm to builtins.genericClosure manually. Also note that
+     using tailCallOpt doesn't mean that the stack won't ever overflow:
+     Data structures, especially lazy ones, can still cause all the
+     available stack space to be consumed.
+
+     The optimization also only concerns avoiding stack overflows,
+     tailCallOpt will make functions slower if anything.
+
+     Type: (F -> F) -> F where F is any tail recursive function.
+
+     Example:
+
+     let
+       label' = self: acc: n:
+         if n == 0
+         then "This is " + acc + "cursed."
+         else self (acc + "very ") (n - 1);
+
+       # Equivalent to a naive recursive implementation in Nix
+       label = (lib.fix label') "";
+
+       labelOpt = (tailCallOpt label') "";
+     in
+
+     label 5
+     => "This is very very very very very cursed."
+
+     labelOpt 5
+     => "This is very very very very very cursed."
+
+     label 10000
+     => error: stack overflow (possible infinite recursion)
+
+     labelOpt 10000
+     => "This is very very very very very very very very very…
+  */
+  tailCallOpt = f:
+    let
+      argc = argCount (lib.fix f);
+
+      # This function simulates being f for f's self reference. Instead of
+      # recursing, it will just return the arguments received as a specially
+      # tagged set, so the recursion step can be performed later.
+      fakef = unapply argc (args: {
+        __tailCall = true;
+        inherit args;
+      });
+      # Pass fakef to f so that it'll be called instead of recursing, ensuring
+      # only one recursion step is performed at a time.
+      encodedf = f fakef;
+
+      opt = args:
+        let
+          steps = builtins.genericClosure {
+            # This is how we encode a (tail) call: A set with final == false
+            # and the list of arguments to pass to be found in args.
+            startSet = [
+              {
+                key = "0";
+                id = 0;
+                final = false;
+                inherit args;
+              }
+            ];
+
+            operator =
+              { id, final, ... }@state:
+              let
+                # Plumbing to make genericClosure happy
+                newIds = {
+                  key = toString (id + 1);
+                  id = id + 1;
+                };
+
+                # Perform recursion step
+                call = apply encodedf state.args;
+
+                # If call encodes a new call, return the new encoded call,
+                # otherwise signal that we're done.
+                newState =
+                  if builtins.isAttrs call && call.__tailCall or false
+                  then newIds // {
+                    final = false;
+                    inherit (call) args;
+                  } else newIds // {
+                    final = true;
+                    value = call;
+                  };
+              in
+
+              if final
+              then [ ] # end condition for genericClosure
+              else [ newState ];
+          };
+        in
+        # The returned list contains intermediate steps we ignore.
+        (builtins.head (builtins.filter (x: x.final) steps)).value;
+    in
+    unapply argc opt;
+in
+
+{
+  inherit (lib)
+    fix
+    flip
+    const
+    ;
+
+  inherit
+    id
+    rl
+    rls
+    lr
+    lrs
+    hasEllipsis
+    argCount
+    tailCallOpt
+    apply
+    unapply
+    ;
+}
diff --git a/users/sterni/nix/fun/tests/default.nix b/users/sterni/nix/fun/tests/default.nix
new file mode 100644
index 0000000000..6b1e6fcc7b
--- /dev/null
+++ b/users/sterni/nix/fun/tests/default.nix
@@ -0,0 +1,82 @@
+{ depot, ... }:
+
+let
+  inherit (depot.nix.runTestsuite)
+    runTestsuite
+    it
+    assertEq
+    ;
+
+  inherit (depot.nix) escapeExecline;
+
+  inherit (depot.users.sterni.nix)
+    fun
+    ;
+
+  hasEllipsisTests = it "checks fun.hasEllipsis" [
+    (assertEq "Malicious string" false
+      (fun.hasEllipsis (builtins.toXML ({ foo, ... }: 12))))
+    (assertEq "No function" false
+      (fun.hasEllipsis 23))
+    (assertEq "No attribute set pattern" false
+      (fun.hasEllipsis (a: a + 2)))
+    (assertEq "No ellipsis" false
+      (fun.hasEllipsis ({ foo, bar }: foo + bar)))
+    (assertEq "Ellipsis" true
+      (fun.hasEllipsis ({ depot, pkgs, ... }: 42)))
+  ];
+
+  argCountTests = it "checks fun.argCount" [
+    (assertEq "builtins.sub has two arguments" 2
+      (fun.argCount builtins.sub))
+    (assertEq "fun.argCount has one argument" 1
+      (fun.argCount fun.argCount))
+    (assertEq "runTestsuite has two arguments" 2
+      (fun.argCount runTestsuite))
+  ];
+
+  applyTests = it "checks that fun.apply is equivalent to calling" [
+    (assertEq "fun.apply builtins.sub" (builtins.sub 23 42)
+      (fun.apply builtins.sub [ 23 42 ]))
+    (assertEq "fun.apply escapeExecline" (escapeExecline [ "foo" [ "bar" ] ])
+      (fun.apply escapeExecline [ [ "foo" [ "bar" ] ] ]))
+  ];
+
+  unapplyTests = it "checks fun.unapply" [
+    (assertEq "fun.unapply 3 accepts 3 args" 3
+      (fun.argCount (fun.unapply 3 fun.id)))
+    (assertEq "fun.unapply 73 accepts 73 args" 73
+      (fun.argCount (fun.unapply 73 fun.id)))
+    (assertEq "fun.unapply 1 accepts 73 args" 1
+      (fun.argCount (fun.unapply 1 fun.id)))
+    (assertEq "fun.unapply collects arguments correctly"
+      (fun.unapply 5 fun.id 1 2 3 4 5)
+      [ 1 2 3 4 5 ])
+    (assertEq "fun.unapply calls the given function correctly" 1
+      (fun.unapply 1 builtins.head 1))
+  ];
+
+  fac' = self: acc: n: if n == 0 then acc else self (n * acc) (n - 1);
+
+  facPlain = fun.fix fac' 1;
+  facOpt = fun.tailCallOpt fac' 1;
+
+  tailCallOptTests = it "checks fun.tailCallOpt" [
+    (assertEq "optimized and unoptimized factorial have the same base case"
+      (facPlain 0)
+      (facOpt 0))
+    (assertEq "optimized and unoptimized factorial have same value for 1"
+      (facPlain 1)
+      (facOpt 1))
+    (assertEq "optimized and unoptimized factorial have same value for 100"
+      (facPlain 100)
+      (facOpt 100))
+  ];
+in
+runTestsuite "nix.fun" [
+  hasEllipsisTests
+  argCountTests
+  applyTests
+  unapplyTests
+  tailCallOptTests
+]
diff --git a/users/sterni/nix/html/README.md b/users/sterni/nix/html/README.md
new file mode 100644
index 0000000000..0349e466a1
--- /dev/null
+++ b/users/sterni/nix/html/README.md
@@ -0,0 +1,148 @@
+# html.nix — _the_ most cursed Nix HTML DSL
+
+A quick example to show you what it looks like:
+
+```nix
+# Note: this example is for standalone usage out of depot
+{ pkgs ? import <nixpkgs> {} }:
+
+let
+  # zero dependency, one file implementation
+  htmlNix = import ./path/to/html.nix { };
+
+  # make the magic work
+  inherit (htmlNix) __findFile esc withDoctype;
+in
+
+pkgs.writeText "example.html" (withDoctype (<html> {} [
+  (<head> {} [
+    (<meta> { charset = "utf-8"; } null)
+    (<title> {} (esc "hello world"))
+  ])
+  (<body> {} [
+    (<h1> {} (esc "hello world"))
+    (<p> { class = "intro"; } (esc ''
+      welcome to the land of sillyness!
+    ''))
+    (<ul> {} [
+      (<li> {} [
+        (esc "check out ")
+        (<a> { href = "https://code.tvl.fyi"; } "depot")
+      ])
+      (<li> {} [
+        (esc "find ")
+        (<a> { href = "https://cl.tvl.fyi/q/hashtag:cursed"; } "cursed things")
+      ])
+    ])
+  ])
+]))
+```
+
+Convince yourself it works:
+
+```console
+$ $BROWSER $(nix-build example.nix)
+```
+
+Alternatively, in depot:
+
+```console
+$ $BROWSER $(nix-build -A users.sterni.nix.html.tests)
+```
+
+## Creating tags
+
+An empty tag is passed `null` as its content argument:
+
+```nix
+<link> {
+  rel = "stylesheet";
+  href = "/main.css";
+  type = "text/css";
+} null
+
+# => "<link href=\"/main.css\" rel=\"stylesheet\" type=\"text/css\"/>"
+```
+
+Content is expected to be HTML:
+
+```nix
+<div> { class = "foo"; } "<strong>hi</strong>"
+
+# => "<div class=\"foo\"><strong>hi</strong></div>"
+```
+
+If it's not, be sure to escape it:
+
+```nix
+<p> {} (esc "A => B")
+
+# => "<p>A =&gt; B</p>"
+```
+
+Nesting tags works of course:
+
+```nix
+<div> {} (<strong> {} (<em> {} "hi"))
+
+# => "<div><strong><em>hi</em></strong></div>"
+```
+
+If the content of a tag is a list, it's concatenated:
+
+```nix
+<h1> {} [
+  (esc "The ")
+  (<strong> {} "Nix")
+  (esc " ")
+  (<em> {} "Expression")
+  (esc " Language")
+]
+
+# => "<h1>The <strong>Nix</strong> <em>Expression</em> Language</h1>"
+```
+
+More detailed documentation can be found in `nixdoc`-compatible
+comments in the source file (`default.nix` in this directory).
+
+## How does this work?
+
+*Theoretically* expressions like `<nixpkgs>` are just ordinary paths —
+their actual value is determined from `NIX_PATH`. `html.nix` works
+because of how this is actually implemented: At [parse time][spath-parsing]
+Nix transparently translates an expression like `<foo>` into
+`__findFile __nixPath "foo"`:
+
+```
+nix-repl> <nixpkgs>
+/nix/var/nix/profiles/per-user/root/channels/vuizvui/nixpkgs
+
+nix-repl> __findFile __nixPath "nixpkgs"
+/nix/var/nix/profiles/per-user/root/channels/vuizvui/nixpkgs
+```
+
+This translation doesn't take any scoping issues into account --
+so we can just shadow `__findFile` and make it return anything,
+even a function:
+
+```
+nix-repl> __findFile = nixPath: str:
+            /**/ if str == "double" then x: x * 2
+            else if str == "triple" then x: x * 3
+            else throw "what?"
+
+nix-repl> <double> 2
+4
+
+nix-repl> <triple> 3
+9
+
+nix-repl> <quadruple> 4
+error: what?
+```
+
+Exactly this is what we are doing in `html.nix`:
+Using `let inherit (htmlNix) __findFile; in` we shadow the builtin `__findFile`
+with a function which returns a function rendering a particular HTML tag.
+
+[spath-parsing]: https://github.com/NixOS/nix/blob/293220bed5a75efc963e33c183787e87e55e28d9/src/libexpr/parser.y#L410-L416
diff --git a/users/sterni/nix/html/default.nix b/users/sterni/nix/html/default.nix
new file mode 100644
index 0000000000..d25a7ab8da
--- /dev/null
+++ b/users/sterni/nix/html/default.nix
@@ -0,0 +1,122 @@
+# Copyright © 2021 sterni
+# SPDX-License-Identifier: MIT
+#
+# This file provides a cursed HTML DSL for nix which works by overloading
+# the NIX_PATH lookup operation via angle bracket operations, e. g. `<nixpkgs>`.
+
+{ ... }:
+
+let
+  /* Escape everything we have to escape in an HTML document if either
+     in a normal context or an attribute string (`<>&"'`).
+
+     A shorthand for this function called `esc` is also provided.
+
+     Type: string -> string
+
+     Example:
+
+     escapeMinimal "<hello>"
+     => "&lt;hello&gt;"
+  */
+  escapeMinimal = builtins.replaceStrings
+    [ "<" ">" "&" "\"" "'" ]
+    [ "&lt;" "&gt;" "&amp;" "&quot;" "&#039;" ];
+
+  /* Return a string with a correctly rendered tag of the given name,
+     with the given attributes which are automatically escaped.
+
+     If the content argument is `null`, the tag will have no children nor a
+     closing element. If the content argument is a string it is used as the
+     content as is (unescaped). If the content argument is a list, its
+     elements are concatenated.
+
+     `renderTag` is only an internal function which is reexposed as `__findFile`
+     to allow for much neater syntax than calling `renderTag` everywhere:
+
+     ```nix
+     { depot, ... }:
+     let
+       inherit (depot.users.sterni.nix.html) __findFile esc;
+     in
+
+     <html> {} [
+       (<head> {} (<title> {} (esc "hello world")))
+       (<body> {} [
+         (<h1> {} (esc "hello world"))
+         (<p> {} (esc "foo bar"))
+       ])
+     ]
+
+     ```
+
+     As you can see, the need to call a function disappears, instead the
+     `NIX_PATH` lookup operation via `<foo>` is overloaded, so it becomes
+     `renderTag "foo"` automatically.
+
+     Since the content argument may contain the result of other `renderTag`
+     calls, we can't escape it automatically. Instead this must be done manually
+     using `esc`.
+
+     Type: string -> attrs<string> -> (list<string> | string | null) -> string
+
+     Example:
+
+     <link> {
+       rel = "stylesheet";
+       href = "/css/main.css";
+       type = "text/css";
+     } null
+
+     renderTag "link" {
+       rel = "stylesheet";
+       href = "/css/main.css";
+       type = "text/css";
+     } null
+
+     => "<link href=\"/css/main.css\" rel=\"stylesheet\" type=\"text/css\"/>"
+
+     <p> {} [
+       "foo "
+       (<strong> {} "bar")
+     ]
+
+     renderTag "p" {} "foo <strong>bar</strong>"
+     => "<p>foo <strong>bar</strong></p>"
+  */
+  renderTag = tag: attrs: content:
+    let
+      attrs' = builtins.concatStringsSep "" (
+        builtins.map
+          (n:
+            " ${escapeMinimal n}=\"${escapeMinimal (toString attrs.${n})}\""
+          )
+          (builtins.attrNames attrs)
+      );
+      content' =
+        if builtins.isList content
+        then builtins.concatStringsSep "" content
+        else content;
+    in
+    if content == null
+    then "<${tag}${attrs'}/>"
+    else "<${tag}${attrs'}>${content'}</${tag}>";
+
+  /* Prepend "<!DOCTYPE html>" to a string.
+
+     Type: string -> string
+
+     Example:
+
+     withDoctype (<body> {} (esc "hello"))
+     => "<!DOCTYPE html><body>hello</body>"
+  */
+  withDoctype = doc: "<!DOCTYPE html>" + doc;
+
+in
+{
+  inherit escapeMinimal renderTag withDoctype;
+
+  __findFile = _: renderTag;
+  esc = escapeMinimal;
+}
diff --git a/users/sterni/nix/html/tests/default.nix b/users/sterni/nix/html/tests/default.nix
new file mode 100644
index 0000000000..0d80f2f1cd
--- /dev/null
+++ b/users/sterni/nix/html/tests/default.nix
@@ -0,0 +1,93 @@
+{ depot, pkgs, ... }:
+
+let
+  inherit (depot.users.sterni.nix.html)
+    __findFile
+    esc
+    withDoctype
+    ;
+
+  exampleDocument = withDoctype (<html> { lang = "en"; } [
+    (<head> { } [
+      (<meta> { charset = "utf-8"; } null)
+      (<title> { } "html.nix example document")
+      (<link>
+        {
+          rel = "license";
+          href = "https://code.tvl.fyi/about/LICENSE";
+          type = "text/html";
+        }
+        null)
+      (<style> { } (esc ''
+        hgroup h2 {
+          font-weight: normal;
+        }
+
+        dd {
+          margin: 0;
+        }
+      ''))
+    ])
+    (<body> { } [
+      (<main> { } [
+        (<hgroup> { } [
+          (<h1> { } (esc "html.nix"))
+          (<h2> { } [
+            (<em> { } "the")
+            (esc " most cursed HTML DSL ever!")
+          ])
+        ])
+        (<dl> { } [
+          (<dt> { } [
+            (esc "Q: Wait, it's all ")
+            (<a>
+              {
+                href = "https://cl.tvl.fyi/q/hashtag:cursed";
+              }
+              (esc "cursed"))
+            (esc " nix hacks?")
+          ])
+          (<dd> { } (esc "A: Always has been. 🔫"))
+          (<dt> { } (esc "Q: Why does this work?"))
+          (<dd> { } [
+            (esc "Because nix ")
+            (<a>
+              {
+                href = "https://github.com/NixOS/nix/blob/293220bed5a75efc963e33c183787e87e55e28d9/src/libexpr/parser.y#L410-L416";
+              }
+              (esc "translates "))
+            (<a>
+              {
+                href = "https://github.com/NixOS/nix/blob/293220bed5a75efc963e33c183787e87e55e28d9/src/libexpr/lexer.l#L100";
+              }
+              (esc "SPATH tokens"))
+            (esc " like ")
+            (<code> { } (esc "<nixpkgs>"))
+            (esc " into calls to ")
+            (<code> { } (esc "__findFile"))
+            (esc " in the ")
+            (<em> { } (esc "current"))
+            (esc " scope.")
+          ])
+        ])
+      ])
+    ])
+  ]);
+in
+
+pkgs.runCommandNoCC "html.nix.html"
+{
+  passAsFile = [ "exampleDocument" ];
+  inherit exampleDocument;
+  nativeBuildInputs = [ pkgs.html5validator ];
+} ''
+  set -x
+  test "${esc "<> && \" \'"}" = "&lt;&gt; &amp;&amp; &quot; &#039;"
+
+  # slow as hell unfortunately
+  html5validator "$exampleDocumentPath"
+
+  mv "$exampleDocumentPath" "$out"
+
+  set +x
+''
diff --git a/users/sterni/nix/int/default.nix b/users/sterni/nix/int/default.nix
new file mode 100644
index 0000000000..54b5596472
--- /dev/null
+++ b/users/sterni/nix/int/default.nix
@@ -0,0 +1,126 @@
+{ depot, lib, ... }:
+
+let
+
+  # TODO(sterni): implement nix.float and figure out which of these
+  #               functions can be split out into a common nix.num
+  #               library.
+
+  inherit (depot.users.sterni.nix)
+    string
+    ;
+
+  inherit (builtins)
+    bitOr
+    bitAnd
+    bitXor
+    mul
+    div
+    add
+    sub
+    ;
+
+  abs = i: if i < 0 then -i else i;
+
+  exp = base: pow:
+    if pow > 0
+    then base * (exp base (pow - 1))
+    else if pow < 0
+    then 1.0 / exp base (abs pow)
+    else 1;
+
+  bitShiftR = bit: count:
+    if count == 0
+    then bit
+    else div (bitShiftR bit (count - 1)) 2;
+
+  bitShiftL = bit: count:
+    if count == 0
+    then bit
+    else 2 * (bitShiftL bit (count - 1));
+
+  hexdigits = "0123456789ABCDEF";
+
+  toHex = int:
+    let
+      go = i:
+        if i == 0
+        then ""
+        else go (bitShiftR i 4)
+          + string.charAt (bitAnd i 15) hexdigits;
+      sign = lib.optionalString (int < 0) "-";
+    in
+    if int == 0
+    then "0"
+    else "${sign}${go (abs int)}";
+
+  fromHexMap = builtins.listToAttrs
+    (lib.imap0 (i: c: { name = c; value = i; })
+      (lib.stringToCharacters hexdigits));
+
+  fromHex = literal:
+    let
+      negative = string.charAt 0 literal == "-";
+      start = if negative then 1 else 0;
+      len = builtins.stringLength literal;
+      # reversed list of all digits
+      digits = builtins.genList
+        (i: string.charAt (len - 1 - i) literal)
+        (len - start);
+      parsed = builtins.foldl'
+        (v: d: {
+          val = v.val + (fromHexMap."${d}" * v.mul);
+          mul = v.mul * 16;
+        })
+        { val = 0; mul = 1; }
+        digits;
+    in
+    if negative
+    then -parsed.val
+    else parsed.val;
+
+  # A nix integer is a 64bit signed integer
+  maxBound = 9223372036854775807;
+
+  # fun fact: -9223372036854775808 is the lower bound
+  # for a nix integer (as you would expect), but you can't
+  # use it as an integer literal or you'll be greeted with:
+  # error: invalid integer '9223372036854775808'
+  # This is because all int literals when parsing are
+  # positive, negative "literals" are positive literals
+  # which are preceded by the arithmetric negation operator.
+  minBound = -9223372036854775807 - 1;
+
+  odd = x: bitAnd x 1 == 1;
+  even = x: bitAnd x 1 == 0;
+
+  # div and mod behave like quot and rem in Haskell,
+  # i. e. they truncate towards 0
+  mod = a: b: let res = a / b; in a - (res * b);
+
+  inRange = a: b: x: x >= a && x <= b;
+
+in
+{
+  inherit
+    maxBound
+    minBound
+    abs
+    exp
+    odd
+    even
+    add
+    sub
+    mul
+    div
+    mod
+    bitShiftR
+    bitShiftL
+    bitOr
+    bitAnd
+    bitXor
+    toHex
+    fromHex
+    inRange
+    ;
+}
diff --git a/users/sterni/nix/int/tests/default.nix b/users/sterni/nix/int/tests/default.nix
new file mode 100644
index 0000000000..8d2263b421
--- /dev/null
+++ b/users/sterni/nix/int/tests/default.nix
@@ -0,0 +1,459 @@
+{ depot, lib, ... }:
+
+let
+
+  inherit (depot.nix.runTestsuite)
+    runTestsuite
+    it
+    assertEq
+    ;
+
+  inherit (depot.users.sterni.nix)
+    int
+    string
+    fun
+    ;
+
+  testBounds = it "checks minBound and maxBound" [
+    # this is gonna blow up in my face because
+    # integer overflow is undefined behavior in
+    # C++, so most likely anything could happen?
+    (assertEq "maxBound is the maxBound" true
+      (int.maxBound + 1 < int.maxBound))
+    (assertEq "minBound is the minBound" true
+      (int.minBound - 1 > int.minBound))
+    (assertEq "maxBound overflows to minBound"
+      (int.maxBound + 1)
+      int.minBound)
+    (assertEq "minBound overflows to maxBound"
+      (int.minBound - 1)
+      int.maxBound)
+  ];
+
+  expectedBytes = [
+    "00"
+    "01"
+    "02"
+    "03"
+    "04"
+    "05"
+    "06"
+    "07"
+    "08"
+    "09"
+    "0A"
+    "0B"
+    "0C"
+    "0D"
+    "0E"
+    "0F"
+    "10"
+    "11"
+    "12"
+    "13"
+    "14"
+    "15"
+    "16"
+    "17"
+    "18"
+    "19"
+    "1A"
+    "1B"
+    "1C"
+    "1D"
+    "1E"
+    "1F"
+    "20"
+    "21"
+    "22"
+    "23"
+    "24"
+    "25"
+    "26"
+    "27"
+    "28"
+    "29"
+    "2A"
+    "2B"
+    "2C"
+    "2D"
+    "2E"
+    "2F"
+    "30"
+    "31"
+    "32"
+    "33"
+    "34"
+    "35"
+    "36"
+    "37"
+    "38"
+    "39"
+    "3A"
+    "3B"
+    "3C"
+    "3D"
+    "3E"
+    "3F"
+    "40"
+    "41"
+    "42"
+    "43"
+    "44"
+    "45"
+    "46"
+    "47"
+    "48"
+    "49"
+    "4A"
+    "4B"
+    "4C"
+    "4D"
+    "4E"
+    "4F"
+    "50"
+    "51"
+    "52"
+    "53"
+    "54"
+    "55"
+    "56"
+    "57"
+    "58"
+    "59"
+    "5A"
+    "5B"
+    "5C"
+    "5D"
+    "5E"
+    "5F"
+    "60"
+    "61"
+    "62"
+    "63"
+    "64"
+    "65"
+    "66"
+    "67"
+    "68"
+    "69"
+    "6A"
+    "6B"
+    "6C"
+    "6D"
+    "6E"
+    "6F"
+    "70"
+    "71"
+    "72"
+    "73"
+    "74"
+    "75"
+    "76"
+    "77"
+    "78"
+    "79"
+    "7A"
+    "7B"
+    "7C"
+    "7D"
+    "7E"
+    "7F"
+    "80"
+    "81"
+    "82"
+    "83"
+    "84"
+    "85"
+    "86"
+    "87"
+    "88"
+    "89"
+    "8A"
+    "8B"
+    "8C"
+    "8D"
+    "8E"
+    "8F"
+    "90"
+    "91"
+    "92"
+    "93"
+    "94"
+    "95"
+    "96"
+    "97"
+    "98"
+    "99"
+    "9A"
+    "9B"
+    "9C"
+    "9D"
+    "9E"
+    "9F"
+    "A0"
+    "A1"
+    "A2"
+    "A3"
+    "A4"
+    "A5"
+    "A6"
+    "A7"
+    "A8"
+    "A9"
+    "AA"
+    "AB"
+    "AC"
+    "AD"
+    "AE"
+    "AF"
+    "B0"
+    "B1"
+    "B2"
+    "B3"
+    "B4"
+    "B5"
+    "B6"
+    "B7"
+    "B8"
+    "B9"
+    "BA"
+    "BB"
+    "BC"
+    "BD"
+    "BE"
+    "BF"
+    "C0"
+    "C1"
+    "C2"
+    "C3"
+    "C4"
+    "C5"
+    "C6"
+    "C7"
+    "C8"
+    "C9"
+    "CA"
+    "CB"
+    "CC"
+    "CD"
+    "CE"
+    "CF"
+    "D0"
+    "D1"
+    "D2"
+    "D3"
+    "D4"
+    "D5"
+    "D6"
+    "D7"
+    "D8"
+    "D9"
+    "DA"
+    "DB"
+    "DC"
+    "DD"
+    "DE"
+    "DF"
+    "E0"
+    "E1"
+    "E2"
+    "E3"
+    "E4"
+    "E5"
+    "E6"
+    "E7"
+    "E8"
+    "E9"
+    "EA"
+    "EB"
+    "EC"
+    "ED"
+    "EE"
+    "EF"
+    "F0"
+    "F1"
+    "F2"
+    "F3"
+    "F4"
+    "F5"
+    "F6"
+    "F7"
+    "F8"
+    "F9"
+    "FA"
+    "FB"
+    "FC"
+    "FD"
+    "FE"
+    "FF"
+  ];
+
+  hexByte = i: string.fit { width = 2; char = "0"; } (int.toHex i);
+
+  hexInts = [
+    { left = 0; right = "0"; }
+    { left = 1; right = "1"; }
+    { left = 11; right = "B"; }
+    { left = 123; right = "7B"; }
+    { left = 9000; right = "2328"; }
+    { left = 2323; right = "913"; }
+    { left = 4096; right = "1000"; }
+    { left = int.maxBound; right = "7FFFFFFFFFFFFFFF"; }
+    { left = int.minBound; right = "-8000000000000000"; }
+  ];
+
+  testHex = it "checks conversion to hex" (lib.flatten [
+    (lib.imap0
+      (i: hex: [
+        (assertEq "hexByte ${toString i} == ${hex}" (hexByte i) hex)
+        (assertEq "${toString i} == fromHex ${hex}" i (int.fromHex hex))
+      ])
+      expectedBytes)
+    (builtins.map
+      ({ left, right }: [
+        (assertEq "toHex ${toString left} == ${right}" (int.toHex left) right)
+        (assertEq "${toString left} == fromHex ${right}" left (int.fromHex right))
+      ])
+      hexInts)
+  ]);
+
+  testBasic = it "checks basic int operations" [
+    (assertEq "122 is even" (int.even 122 && !(int.odd 122)) true)
+    (assertEq "123 is odd" (int.odd 123 && !(int.even 123)) true)
+    (assertEq "abs -4959" (int.abs (-4959)) 4959)
+  ];
+
+  expNumbers = [
+    { left = -3; right = 0.125; }
+    { left = -2; right = 0.25; }
+    { left = -1; right = 0.5; }
+    { left = 0; right = 1; }
+    { left = 1; right = 2; }
+    { left = 2; right = 4; }
+    { left = 3; right = 8; }
+    { left = 4; right = 16; }
+    { left = 5; right = 32; }
+    { left = 16; right = 65536; }
+  ];
+
+  testExp = it "checks exponentiation"
+    (builtins.map
+      ({ left, right }:
+        assertEq
+          "2 ^ ${toString left} == ${toString right}"
+          (int.exp 2 left)
+          right)
+      expNumbers);
+
+  shifts = [
+    { a = 2; b = 5; c = 64; op = "<<"; }
+    { a = -2; b = 5; c = -64; op = "<<"; }
+    { a = 123; b = 4; c = 1968; op = "<<"; }
+    { a = 1; b = 8; c = 256; op = "<<"; }
+    { a = 256; b = 8; c = 1; op = ">>"; }
+    { a = 374; b = 2; c = 93; op = ">>"; }
+    { a = 2; b = 2; c = 0; op = ">>"; }
+    { a = 99; b = 9; c = 0; op = ">>"; }
+  ];
+
+  checkShift = { a, b, c, op }@args:
+    let
+      f = string.match op {
+        "<<" = int.bitShiftL;
+        ">>" = int.bitShiftR;
+      };
+    in
+    assertEq "${toString a} ${op} ${toString b} == ${toString c}" (f a b) c;
+
+  checkShiftRDivExp = n:
+    assertEq "${toString n} >> 5 == ${toString n} / 2 ^ 5"
+      (int.bitShiftR n 5)
+      (int.div n (int.exp 2 5));
+
+  checkShiftLMulExp = n:
+    assertEq "${toString n} >> 6 == ${toString n} * 2 ^ 6"
+      (int.bitShiftL n 5)
+      (int.mul n (int.exp 2 5));
+
+  testBit = it "checks bitwise operations" (lib.flatten [
+    (builtins.map checkShift shifts)
+    (builtins.map checkShiftRDivExp [
+      1
+      2
+      3
+      5
+      7
+      23
+      1623
+      238
+      34
+      348
+      2834
+      834
+      348
+    ])
+    (builtins.map checkShiftLMulExp [
+      1
+      2
+      3
+      5
+      7
+      23
+      384
+      3
+      2
+      5991
+      85109
+      38
+    ])
+  ]);
+
+  divisions = [
+    { a = 2; b = 1; c = 2; mod = 0; }
+    { a = 2; b = 2; c = 1; mod = 0; }
+    { a = 20; b = 10; c = 2; mod = 0; }
+    { a = 12; b = 5; c = 2; mod = 2; }
+    { a = 23; b = 4; c = 5; mod = 3; }
+  ];
+
+  checkDiv = n: { a, b, c, mod }: [
+    (assertEq "${n}: div result" (int.div a b) c)
+    (assertEq "${n}: mod result" (int.mod a b) mod)
+    (assertEq "${n}: divMod law" ((int.div a b) * b + (int.mod a b)) a)
+  ];
+
+  testDivMod = it "checks integer division and modulo"
+    (lib.flatten [
+      (builtins.map (checkDiv "+a / +b") divisions)
+      (builtins.map
+        (fun.rl (checkDiv "-a / +b") (x: x // {
+          a = -x.a;
+          c = -x.c;
+          mod = -x.mod;
+        }))
+        divisions)
+      (builtins.map
+        (fun.rl (checkDiv "+a / -b") (x: x // {
+          b = -x.b;
+          c = -x.c;
+        }))
+        divisions)
+      (builtins.map
+        (fun.rl (checkDiv "-a / -b") (x: x // {
+          a = -x.a;
+          b = -x.b;
+          mod = -x.mod;
+        }))
+        divisions)
+    ]);
+
+in
+runTestsuite "nix.int" [
+  testBounds
+  testHex
+  testBasic
+  testExp
+  testBit
+  testDivMod
+]
diff --git a/users/sterni/nix/string/default.nix b/users/sterni/nix/string/default.nix
new file mode 100644
index 0000000000..852ef2538f
--- /dev/null
+++ b/users/sterni/nix/string/default.nix
@@ -0,0 +1,122 @@
+{ depot, lib, ... }:
+
+let
+
+  inherit (depot.users.sterni.nix.char)
+    chr
+    ord
+    ;
+
+  inherit (depot.users.sterni.nix)
+    int
+    flow
+    ;
+
+  take = n: s:
+    builtins.substring 0 n s;
+
+  drop = n: s:
+    builtins.substring n int.maxBound s;
+
+  charAt = i: s:
+    let
+      r = builtins.substring i 1 s;
+    in
+    if r == "" then null else r;
+
+  charIndex = char: s:
+    let
+      len = builtins.stringLength s;
+      go = i:
+        flow.cond [
+          [ (i >= len) null ]
+          [ (charAt i s == char) i ]
+          [ true (go (i + 1)) ]
+        ];
+    in
+    go 0;
+
+  toChars = lib.stringToCharacters;
+  fromChars = lib.concatStrings;
+
+  toBytes = str:
+    builtins.map ord (toChars str);
+
+  fromBytes = is: lib.concatMapStrings chr is;
+
+  pad = { left ? 0, right ? 0, char ? " " }: s:
+    let
+      leftS = fromChars (builtins.genList (_: char) left);
+      rightS = fromChars (builtins.genList (_: char) right);
+    in
+    "${leftS}${s}${rightS}";
+
+  fit = { char ? " ", width, side ? "left" }: s:
+    let
+      diff = width - builtins.stringLength s;
+    in
+    if diff <= 0
+    then s
+    else pad { inherit char; "${side}" = diff; } s;
+
+  # pattern matching for strings only
+  match = val: matcher: matcher."${val}";
+
+  /* Bare-bones printf implementation. Supported format specifiers:
+
+     * `%%` escapes `%`
+     * `%s` is substituted by a string
+
+     As expected, the first argument is a format string and the values
+     for its format specifiers need to provided as the next arguments
+     in order.
+
+     Type: string -> (printfVal : either string (a -> printfVal))
+  */
+  printf = formatString:
+    let
+      specifierWithArg = token: builtins.elem token [
+        "%s"
+      ];
+      isSpecifier = lib.hasPrefix "%";
+
+      tokens = lib.flatten (builtins.split "(%.)" formatString);
+      argsNeeded = builtins.length (builtins.filter specifierWithArg tokens);
+
+      format = args: (builtins.foldl'
+        ({ out ? "", argIndex ? 0 }: token: {
+          argIndex = argIndex + (if specifierWithArg token then 1 else 0);
+          out =
+            /**/
+            if token == "%s" then out + builtins.elemAt args argIndex
+            else if token == "%%" then out + "%"
+            else if isSpecifier token then throw "Unsupported format specifier ${token}"
+            else out + token;
+        })
+        { }
+        tokens).out;
+
+      accumulateArgs = argCount: args:
+        if argCount > 0
+        then arg: accumulateArgs (argCount - 1) (args ++ [ arg ])
+        else format args;
+    in
+    accumulateArgs argsNeeded [ ];
+
+in
+{
+  inherit
+    take
+    drop
+    charAt
+    charIndex
+    toBytes
+    fromBytes
+    toChars
+    fromChars
+    pad
+    fit
+    match
+    printf
+    ;
+}
diff --git a/users/sterni/nix/string/tests/default.nix b/users/sterni/nix/string/tests/default.nix
new file mode 100644
index 0000000000..e9015e95dc
--- /dev/null
+++ b/users/sterni/nix/string/tests/default.nix
@@ -0,0 +1,72 @@
+{ depot, ... }:
+
+let
+
+  inherit (depot.users.sterni.nix)
+    string
+    ;
+
+  inherit (depot.nix.runTestsuite)
+    it
+    assertEq
+    runTestsuite
+    ;
+
+  testTakeDrop = it "tests take and drop" [
+    (assertEq "take"
+      (string.take 5 "five and more")
+      "five ")
+    (assertEq "drop"
+      (string.drop 2 "coin")
+      "in")
+    (assertEq "take out of bounds"
+      (string.take 100 "foo")
+      "foo")
+    (assertEq "drop out of bounds"
+      (string.drop 42 "lol")
+      "")
+  ];
+
+  testIndexing = it "tests string indexing" [
+    (assertEq "normal charAt"
+      (string.charAt 3 "helo")
+      "o")
+    (assertEq "out of bounds charAt"
+      (string.charAt 5 "helo")
+      null)
+  ];
+
+  testFinding = it "tests finding in strings" [
+    (assertEq "normal charIndex"
+      (string.charIndex "d" "abcdefghijkl")
+      3)
+    (assertEq "charIndex no match"
+      (string.charIndex "w" "zZzZzzzZZZ")
+      null)
+  ];
+
+  dontEval = builtins.throw "this should not get evaluated";
+
+  testMatch = it "tests match" [
+    (assertEq "basic match usage" 42
+      (string.match "answer" {
+        "answer" = 42;
+        "banana" = dontEval;
+        "maleur" = dontEval;
+      }))
+  ];
+
+  f = "f";
+  testPrintf = it "prints f" [
+    (assertEq "basic %s usage" "print ${f}" (string.printf "print %s" f))
+    (assertEq "% escaping" "100%" (string.printf "100%%"))
+  ];
+
+in
+runTestsuite "nix.string" [
+  testTakeDrop
+  testIndexing
+  testFinding
+  testMatch
+  testPrintf
+]
diff --git a/users/sterni/nix/url/default.nix b/users/sterni/nix/url/default.nix
new file mode 100644
index 0000000000..4a401873a1
--- /dev/null
+++ b/users/sterni/nix/url/default.nix
@@ -0,0 +1,100 @@
+{ depot, lib, ... }:
+
+let
+
+  inherit (depot.users.sterni.nix)
+    char
+    int
+    string
+    flow
+    ;
+
+  reserved = c: builtins.elem c [
+    "!"
+    "#"
+    "$"
+    "&"
+    "'"
+    "("
+    ")"
+    "*"
+    "+"
+    ","
+    "/"
+    ":"
+    ";"
+    "="
+    "?"
+    "@"
+    "["
+    "]"
+  ];
+
+  unreserved = c: char.asciiAlphaNum c
+    || builtins.elem c [ "-" "_" "." "~" ];
+
+  percentEncode = c:
+    if unreserved c
+    then c
+    else "%" + (string.fit
+      {
+        width = 2;
+        char = "0";
+        side = "left";
+      }
+      (int.toHex (char.ord c)));
+
+  encode = { leaveReserved ? false }: s:
+    let
+      chars = lib.stringToCharacters s;
+      tr = c:
+        if leaveReserved && reserved c
+        then c
+        else percentEncode c;
+    in
+    lib.concatStrings (builtins.map tr chars);
+
+  decode = s:
+    let
+      tokens = builtins.split "%" s;
+      decodeStep =
+        { result ? ""
+        , inPercent ? false
+        }: s:
+        flow.cond [
+          [
+            (builtins.isList s)
+            {
+              inherit result;
+              inPercent = true;
+            }
+          ]
+          [
+            inPercent
+            {
+              inPercent = false;
+              # first two characters came after an %
+              # the rest is the string until the next %
+              result = result
+                + char.chr (int.fromHex (string.take 2 s))
+                + (string.drop 2 s);
+            }
+          ]
+          [
+            (!inPercent)
+            {
+              result = result + s;
+            }
+          ]
+        ];
+
+    in
+    (builtins.foldl' decodeStep { } tokens).result;
+
+in
+{
+  inherit
+    encode
+    decode
+    ;
+}
diff --git a/users/sterni/nix/url/tests/default.nix b/users/sterni/nix/url/tests/default.nix
new file mode 100644
index 0000000000..4eb6f95ccd
--- /dev/null
+++ b/users/sterni/nix/url/tests/default.nix
@@ -0,0 +1,58 @@
+{ depot, ... }:
+
+let
+
+  inherit (depot.nix.runTestsuite)
+    it
+    assertEq
+    runTestsuite
+    ;
+
+  inherit (depot.users.sterni.nix)
+    url
+    ;
+
+  checkEncoding = args: { left, right }:
+    assertEq "encode ${builtins.toJSON left} == ${builtins.toJSON right}"
+      (url.encode args left)
+      right;
+
+  checkDecoding = { left, right }:
+    assertEq "${builtins.toJSON left} == decode ${builtins.toJSON right}"
+      (url.decode left)
+      right;
+
+  unreserved = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789.-_~";
+
+  encodeExpected = [
+    { left = "Laguna Beach"; right = "Laguna%20Beach"; }
+    { left = "👾 Exterminate!"; right = "%F0%9F%91%BE%20Exterminate%21"; }
+    { left = unreserved; right = unreserved; }
+    {
+      left = "`!@#$%^&*()+={}[]:;'\\|<>,?/ \"";
+      right = "%60%21%40%23%24%25%5E%26%2A%28%29%2B%3D%7B%7D%5B%5D%3A%3B%27%5C%7C%3C%3E%2C%3F%2F%20%22";
+    }
+  ];
+
+  testEncode = it "checks url.encode"
+    (builtins.map (checkEncoding { }) encodeExpected);
+
+  testDecode = it "checks url.decode"
+    (builtins.map checkDecoding encodeExpected);
+
+  testLeaveReserved = it "checks that leaveReserved is like id for valid URLs"
+    (builtins.map (x: checkEncoding { leaveReserved = true; } { left = x; right = x; }) [
+      "ftp://ftp.is.co.za/rfc/rfc1808.txt"
+      "http://www.ietf.org/rfc/rfc2396.txt"
+      "ldap://[2001:db8::7]/c=GB?objectClass?one"
+      "mailto:John.Doe@example.com"
+      "news:comp.infosystems.www.servers.unix"
+      "tel:+1-816-555-1212"
+      "telnet://192.0.2.16:80/"
+      "urn:oasis:names:specification:docbook:dtd:xml:4.1.2"
+    ]);
+in
+runTestsuite "nix.url" [
+  testEncode
+  testLeaveReserved
+]
diff --git a/users/sterni/nix/utf8/default.nix b/users/sterni/nix/utf8/default.nix
new file mode 100644
index 0000000000..71c846c042
--- /dev/null
+++ b/users/sterni/nix/utf8/default.nix
@@ -0,0 +1,325 @@
+{ depot, lib, ... }:
+
+let
+
+  inherit (depot.users.sterni.nix)
+    char
+    flow
+    fun
+    int
+    string
+    util
+    ;
+
+  /* (Internal) function to determine the amount
+     bytes left in a UTF-8 byte sequence from the
+     first byte.
+
+     This function will throw if the given first
+     byte is ill-formed, but will not detect all
+     cases of ill-formed-ness.
+
+     Based on table 3-6. from The Unicode Standard,
+     Version 13.0, section 3.9.
+
+     Type: integer -> integer
+  */
+  byteCount = i: flow.cond [
+    [ (int.bitAnd i 128 == 0) 1 ]
+    [ (int.bitAnd i 224 == 192) 2 ]
+    [ (int.bitAnd i 240 == 224) 3 ]
+    [ (int.bitAnd i 248 == 240) 4 ]
+    [ true (builtins.throw "Ill-formed first byte ${int.toHex i}") ]
+  ];
+
+  /* (Internal) function to check if a given byte in
+     an UTF-8 byte sequence is well-formed.
+
+     Based on table 3-7. from The Unicode Standard,
+     Version 13.0, section 3.9.
+
+     Type: integer -> integer -> integer -> bool
+  */
+  wellFormedByte =
+    # first byte's integer value
+    first:
+    # byte position as an index starting with 0
+    pos:
+    let
+      defaultRange = int.inRange 128 191;
+
+      secondBytePredicate = flow.switch first [
+        [ (int.inRange 194 223) defaultRange ] # C2..DF
+        [ 224 (int.inRange 160 191) ] # E0
+        [ (int.inRange 225 236) defaultRange ] # E1..EC
+        [ 237 (int.inRange 128 159) ] # ED
+        [ (int.inRange 238 239) defaultRange ] # EE..EF
+        [ 240 (int.inRange 144 191) ] # F0
+        [ (int.inRange 241 243) defaultRange ] # F1..F3
+        [ 244 (int.inRange 128 143) ] # F4
+        [ (fun.const true) null ]
+      ];
+
+      firstBytePredicate = byte: assert first == byte;
+        first < 128 || secondBytePredicate != null;
+    in
+    # Either ASCII or in one of the byte ranges of Table 3-6.
+    if pos == 0 then firstBytePredicate
+    # return predicate according to Table 3-6.
+    else if pos == 1 then assert secondBytePredicate != null; secondBytePredicate
+    # 3rd and 4th byte have only one validity rule
+    else defaultRange;
+
+  /* Iteration step for decoding an UTF-8 byte sequence.
+     It decodes incrementally, i. e. it has to be fed
+     one byte at a time and then returns either a
+     new state or a final result.
+
+     If the resulting attribute set contains the attribute
+     result, it is finished and the decoded codepoint is
+     contained in that attribute. In all other cases,
+     pass the returned set to step again along with
+     a new byte. The initial state to pass is the empty
+     set.
+
+     Extra attributes are always passed through, so you
+     can pass extra state. Be sure not to use result,
+     pos, code, first or count.
+
+     This function will throw with a fairly detailed
+     message if it encounters ill-formed bytes.
+
+     The implementation is based on The Unicode Standard,
+     Version 13.0, section 3.9, especially table 3-6.
+
+     Type: { ... } -> string -> ({ result :: integer, ... } | { ... })
+
+     Example: utf8.step {} "f"
+     => { result = 102; }
+  */
+  step = { pos ? 0, code ? 0, ... }@args: byte:
+    let
+      value = char.ord byte;
+      # first byte is context for well-formed-ness
+      first = args.first or value;
+      count = args.count or (byteCount first);
+      newCode =
+        if count == 1
+        then int.bitAnd 127 first # ascii character
+        else # multi byte UTF-8 sequence
+          let
+            # Calculate the bitmask for extracting the
+            # codepoint data in the current byte.
+            # If the codepoint is not ASCII, the bits
+            # used for codepoint data differ depending
+            # on the byte position and overall byte
+            # count. The first byte always ignores
+            # the (count + 1) most significant bits.
+            # For all subsequent bytes, the 2 most
+            # significant bits need to be ignored.
+            # See also table 3-6.
+            mask =
+              if pos == 0
+              then int.exp 2 (8 - (count + 1)) - 1
+              else 63;
+            # UTF-8 uses the 6 least significant bits in all
+            # subsequent bytes after the first one. Therefore
+            # We can determine the amount we need to shift
+            # the current value by the amount of bytes left.
+            offset = (count - (pos + 1)) * 6;
+          in
+          code + (int.bitShiftL (int.bitAnd mask value) offset);
+      illFormedMsg =
+        "Ill-formed byte ${int.toHex value} at position ${toString pos} in ${toString count} byte UTF-8 sequence";
+    in
+    if !(wellFormedByte first pos value) then builtins.throw illFormedMsg
+    else if pos + 1 == count
+    then (builtins.removeAttrs args [
+      # allow extra state being passed through
+      "count"
+      "code"
+      "pos"
+      "first"
+    ]) // { result = newCode; }
+    else (builtins.removeAttrs args [ "result" ]) // {
+      inherit count first;
+      code = newCode;
+      pos = pos + 1;
+    };
+
+  /* Decode an UTF-8 string into a list of codepoints.
+
+     Throws if the string is ill-formed UTF-8.
+
+     Type: string -> [ integer ]
+  */
+  # TODO(sterni): option to fallback to replacement char instead of failure
+  decode = s:
+    let
+      stringLength = builtins.stringLength s;
+      iterResult = builtins.genericClosure {
+        startSet = [
+          {
+            key = "start";
+            stringIndex = -1;
+            state = { };
+            codepoint = null;
+          }
+        ];
+        operator = { state, stringIndex, ... }:
+          let
+            # updated values for current iteration step
+            newIndex = stringIndex + 1;
+            newState = step state (builtins.substring newIndex 1 s);
+          in
+          lib.optional (newIndex < stringLength) {
+            # unique keys to make genericClosure happy
+            key = toString newIndex;
+            # carryover state for the next step
+            stringIndex = newIndex;
+            state = newState;
+            # actual payload for later, steps with value null are filtered out
+            codepoint = newState.result or null;
+          };
+      };
+    in
+    # extract all steps that yield a code point into a list
+    builtins.map (v: v.codepoint) (
+      builtins.filter
+        (
+          { codepoint, stringIndex, state, ... }:
+
+          let
+            # error message in case we are missing bytes at the end of input
+            earlyEndMsg =
+              if state ? count && state ? pos
+              then "Missing ${toString (with state; count - pos)} bytes at end of input"
+              else "Unexpected end of input";
+          in
+
+          # filter out all iteration steps without a codepoint value
+          codepoint != null
+          # if we are at the iteration step of a non-empty input string, throw
+          # an error if no codepoint was returned, as it indicates an incomplete
+          # UTF-8 sequence.
+          || (stringLength > 0 && stringIndex == stringLength - 1 && throw earlyEndMsg)
+
+        )
+        iterResult
+    );
+
+  /* Pretty prints a Unicode codepoint in the U+<HEX> notation.
+
+     Type: integer -> string
+  */
+  formatCodepoint = cp: "U+" + string.fit
+    {
+      width = 4;
+      char = "0";
+    }
+    (int.toHex cp);
+
+  encodeCodepoint = cp:
+    let
+      # Find the amount of bytes needed to encode the given codepoint.
+      # Note that this doesn't check if the Unicode codepoint is allowed,
+      # but rather allows all theoretically UTF-8-encodeable ones.
+      count = flow.switch cp [
+        [ (int.inRange 0 127) 1 ] # 00000000 0xxxxxxx
+        [ (int.inRange 128 2047) 2 ] # 00000yyy yyxxxxxx
+        [ (int.inRange 2048 65535) 3 ] # zzzzyyyy yyxxxxxx
+        [ (int.inRange 65536 1114111) 4 ] # 000uuuuu zzzzyyyy yyxxxxxx,
+        # capped at U+10FFFF
+
+        [ (fun.const true) (builtins.throw invalidCodepointMsg) ]
+      ];
+
+      invalidCodepointMsg = "${formatCodepoint cp} is not a Unicode codepoint";
+
+      # Extract the bit ranges x, y, z and u from the given codepoint
+      # according to Table 3-6. from The Unicode Standard, Version 13.0,
+      # section 3.9. u is split into uh and ul since they are used in
+      # different bytes in the end.
+      components = lib.mapAttrs
+        (_: { mask, offset }:
+          int.bitAnd (int.bitShiftR cp offset) mask
+        )
+        {
+          x = {
+            mask = if count > 1 then 63 else 127;
+            offset = 0;
+          };
+          y = {
+            mask = if count > 2 then 63 else 31;
+            offset = 6;
+          };
+          z = {
+            mask = 15;
+            offset = 12;
+          };
+          # u which belongs into the second byte
+          ul = {
+            mask = 3;
+            offset = 16;
+          };
+          # u which belongs into the first byte
+          uh = {
+            mask = 7;
+            offset = 18;
+          };
+        };
+      inherit (components) x y z ul uh;
+
+      # Finally construct the byte sequence for the given codepoint. This is
+      # usually done by using the component and adding a few bits as a prefix
+      # which depends on the length of the sequence. The longer the sequence,
+      # the further back each component is pushed. To simplify this, we
+      # always construct a 4 element list and take the last `count` elements.
+      # Thanks to laziness the bogus values created by this are never evaluated.
+      #
+      # Based on table 3-6. from The Unicode Standard,
+      # Version 13.0, section 3.9.
+      bytes = lib.sublist (4 - count) count [
+        # 11110uuu
+        (uh + 240)
+        # 10uuzzzz or 1110zzzz
+        (z + (if count > 3 then 128 + int.bitShiftL ul 4 else 224))
+        # 10yyyyyy or 110yyyyy
+        (y + (if count > 2 then 128 else 192))
+        # 10xxxxxx or 0xxxxxxx
+        (x + (if count > 1 then 128 else 0))
+      ];
+
+      firstByte = builtins.head bytes;
+
+      unableToEncodeMessage = "Can't encode ${formatCodepoint cp} as UTF-8";
+
+    in
+    string.fromBytes (
+      builtins.genList
+        (i:
+          let
+            byte = builtins.elemAt bytes i;
+          in
+          if wellFormedByte firstByte i byte
+          then byte
+          else builtins.throw unableToEncodeMessage
+        )
+        count
+    );
+
+  /* Encode a list of Unicode codepoints into an UTF-8 string.
+
+     Type: [ integer ] -> string
+  */
+  encode = lib.concatMapStrings encodeCodepoint;
+
+in
+{
+  inherit
+    encode
+    decode
+    step
+    formatCodepoint
+    ;
+}
diff --git a/users/sterni/nix/utf8/tests/default.nix b/users/sterni/nix/utf8/tests/default.nix
new file mode 100644
index 0000000000..40783eab24
--- /dev/null
+++ b/users/sterni/nix/utf8/tests/default.nix
@@ -0,0 +1,148 @@
+{ depot, pkgs, lib, ... }:
+
+let
+
+  inherit (pkgs)
+    runCommandLocal
+    ;
+
+  inherit (depot.nix.runTestsuite)
+    runTestsuite
+    it
+    assertEq
+    assertThrows
+    assertDoesNotThrow
+    ;
+
+  inherit (depot.nix.writers)
+    rustSimple
+    ;
+
+  inherit (depot.users.sterni.nix)
+    int
+    utf8
+    string
+    char
+    ;
+
+  rustDecoder = rustSimple
+    {
+      name = "utf8-decode";
+    } ''
+    use std::io::{self, Read};
+    fn main() -> std::io::Result<()> {
+      let mut buffer = String::new();
+      io::stdin().read_to_string(&mut buffer)?;
+
+      print!("[ ");
+
+      for c in buffer.chars() {
+        print!("{} ", u32::from(c));
+      }
+
+      print!("]");
+
+      Ok(())
+    }
+  '';
+
+  rustDecode = s:
+    let
+      expr = runCommandLocal "${s}-decoded" { } ''
+        printf '%s' ${lib.escapeShellArg s} | ${rustDecoder} > $out
+      '';
+    in
+    import expr;
+
+  hexDecode = l:
+    utf8.decode (string.fromBytes (builtins.map int.fromHex l));
+
+  hexEncode = l: utf8.encode (builtins.map int.fromHex l);
+
+  testFailures = it "checks UTF-8 decoding failures" ([
+    (assertThrows "truncated UTF-8 string throws" (hexDecode [ "F0" "9F" ]))
+    # examples from The Unicode Standard
+    (assertThrows "ill-formed: C0 AF" (hexDecode [ "C0" "AF" ]))
+    (assertThrows "ill-formed: E0 9F 80" (hexDecode [ "E0" "9F" "80" ]))
+    (assertEq "well-formed: F4 80 83 92" (hexDecode [ "F4" "80" "83" "92" ]) [ 1048786 ])
+    (assertThrows "Codepoint out of range: 0xFFFFFF" (hexEncode [ "FFFFFF" ]))
+    (assertThrows "Codepoint out of range: -0x02" (hexEncode [ "-02" ]))
+  ] ++ builtins.genList
+    (i:
+      let
+        cp = i + int.fromHex "D800";
+      in
+      assertThrows "Can't encode UTF-16 reserved characters: ${utf8.formatCodepoint cp}"
+        (utf8.encode [ cp ])
+    )
+    (int.fromHex "07FF"));
+
+  testAscii = it "checks decoding of ascii strings"
+    (builtins.map
+      (s: assertEq "ASCII decoding is equal to UTF-8 decoding for \"${s}\""
+        (string.toBytes s)
+        (utf8.decode s)) [
+      "foo bar"
+      "hello\nworld"
+      "carriage\r\nreturn"
+      "1238398494829304 []<><>({})[]!!)"
+      (string.take 127 char.allChars)
+    ]);
+
+  randomUnicode = [
+    "" # empty string should yield empty list
+    "🥰👨‍👨‍👧‍👦🐈‍⬛👩🏽‍🦰"
+    # https://kermitproject.org/utf8.html
+    "ᚠᛇᚻ᛫ᛒᛦᚦ᛫ᚠᚱᚩᚠᚢᚱ᛫ᚠᛁᚱᚪ᛫ᚷᛖᚻᚹᛦᛚᚳᚢᛗ"
+    "An preost wes on leoden, Laȝamon was ihoten"
+    "Sîne klâwen durh die wolken sint geslagen,"
+    "Τὴ γλῶσσα μοῦ ἔδωσαν ἑλληνικὴ"
+    "На берегу пустынных волн"
+    "ვეპხის ტყაოსანი შოთა რუსთაველი"
+    "யாமறிந்த மொழிகளிலே தமிழ்மொழி போல் இனிதாவது எங்கும் காணோம், "
+    "ಬಾ ಇಲ್ಲಿ ಸಂಭವಿಸು "
+  ];
+
+  # https://kermitproject.org/utf8.html
+  glassSentences = [
+    "Euro Symbol: €."
+    "Greek: Μπορώ να φάω σπασμένα γυαλιά χωρίς να πάθω τίποτα."
+    "Íslenska / Icelandic: Ég get etið gler án þess að meiða mig."
+    "Polish: Mogę jeść szkło, i mi nie szkodzi."
+    "Romanian: Pot să mănânc sticlă și ea nu mă rănește."
+    "Ukrainian: Я можу їсти шкло, й воно мені не пошкодить."
+    "Armenian: Կրնամ ապակի ուտել և ինծի անհանգիստ չըներ։"
+    "Georgian: მინას ვჭამ და არა მტკივა."
+    "Hindi: मैं काँच खा सकता हूँ, मुझे उस से कोई पीडा नहीं होती."
+    "Hebrew(2): אני יכול לאכול זכוכית וזה לא מזיק לי."
+    "Yiddish(2): איך קען עסן גלאָז און עס טוט מיר נישט װײ."
+    "Arabic(2): أنا قادر على أكل الزجاج و هذا لا يؤلمني."
+    "Japanese: 私はガラスを食べられます。それは私を傷つけません。"
+    "Thai: ฉันกินกระจกได้ แต่มันไม่ทำให้ฉันเจ็บ "
+  ];
+
+  testDecoding = it "checks decoding of UTF-8 strings against Rust's String"
+    (builtins.map
+      (s: assertEq "Decoding of “${s}” is correct" (utf8.decode s) (rustDecode s))
+      (lib.flatten [
+        glassSentences
+        randomUnicode
+      ]));
+
+  testDecodingEncoding = it "checks that decoding and then encoding forms an identity"
+    (builtins.map
+      (s: assertEq "Decoding and then encoding “${s}” yields itself"
+        (utf8.encode (utf8.decode s))
+        s)
+      (lib.flatten [
+        glassSentences
+        randomUnicode
+      ]));
+
+in
+runTestsuite "nix.utf8" [
+  testFailures
+  testAscii
+  testDecoding
+  testDecodingEncoding
+]