7## u3: noun processing in C
9`u3` is the C library that makes Urbit work. If it wasn't called
10`u3`, it might be called `libnoun` - it's a library for making
13What's a noun? A noun is either a cell or an atom. A cell is an
14ordered pair of any two nouns. An atom is an unsigned integer of
17To the C programmer, this is not a terribly complicated data
18structure, so why do you need a library for it?
20One: nouns have a well-defined computation kernel, Nock, whose
21spec fits on a page and gzips to 340 bytes. But the only
22arithmetic operation in Nock is increment. So it's nontrivial
23to compute both efficiently and correctly.
25Two: `u3` is designed to support "permanent computing," ie, a
26single-level store which is transparently snapshotted. This
27implies a specialized memory-management model, etc, etc.
29(Does `u3` depend on the higher levels of Urbit, Arvo and Hoon?
30Yes and no. `u3` expects you to load something shaped like an
31Arvo kernel, and use it as an event-processing function. But you
32don't need to use this feature if you don't want, and your kernel
33doesn't have to be Arvo proper - just Arvo-compatible. Think of
34`u3` as the BIOS and Arvo as the boot kernel. And there are no
35dependencies at all between Hoon the language and `u3`.)
39Under `u3` is the simple `c3` layer, which is just how we write C
42When writing C in u3, please of course follow the conventions of
43the code around you as regards indentation, etc. It's especially
44important that every function have a header comment, even if it
45says nothing interesting.
47But some of our idiosyncrasies go beyond convention. Yes, we've
48done awful things to C. Here's what we did and why we did.
52First, it's generally acknowledged that underspecified integer
53types are C's worst disaster. C99 fixed this, but the `stdint`
54types are wordy and annoying. We've replaced them with:
58 typedef uint64_t c3_d; // double-word
59 typedef int64_t c3_ds; // signed double-word
60 typedef uint32_t c3_w; // word
61 typedef int32_t c3_ws; // signed word
62 typedef uint16_t c3_s; // short
63 typedef int16_t c3_ss; // signed short
64 typedef uint8_t c3_y; // byte
65 typedef int8_t c3_ys; // signed byte
66 typedef uint8_t c3_b; // bit
68 typedef uint8_t c3_t; // boolean
69 typedef uint8_t c3_o; // loobean
70 typedef uint8_t c3_g; // 5-bit atom for a 32-bit log.
71 typedef uint32_t c3_l; // little; 31-bit unsigned integer
72 typedef uint32_t c3_m; // mote; also c3_l; LSB first a-z 4-char string.
76 typedef char c3_c; // does not match int8_t or uint8_t
77 typedef int c3_i; // int - really bad
78 typedef uintptr_t c3_p; // pointer-length uint - really really bad
79 typedef intptr_t c3_ps; // pointer-length int - really really bad
81Some of these need explanation. A loobean is a Nock boolean -
82Nock, for mysterious reasons, uses 0 as true (always say "yes")
83and 1 as false (always say "no").
85Nock and/or Hoon cannot tell the difference between a short atom
86and a long one, but at the `u3` level every atom under `2^31` is
87direct. The `c3_l` type is useful to annotate this. A `c3_m` is
88a *mote* - a string of up to 4 characters in a `c3_l`, least
89significant byte first. A `c3_g` should be a 5-bit atom. Of
90course, C cannot enforce these constraints, only document them.
92Use the "bad" - ie, poorly specified - integer types only when
93interfacing with external code that expects them.
95An enormous number of motes are defined in `i/c/motes.h`. There
96is no reason to delete motes that aren't being used, or even to
97modularize the definitions. Keep them alphabetical, though.
99### c3: variables and variable naming
101The C3 style uses Hoon style TLV variable names, with a quasi
102Hungarian syntax. This is weird, but works really well, as long
103as what you're doing isn't hideously complicated. (Then it works
104badly, but we shouldn't need anything hideous in u3.)
106A TLV variable name is a random pronounceable three-letter
107string, sometimes with some vague relationship to its meaning,
108but usually not. Usually CVC (consonant-vowel-consonant) is a
111You should use TLVs much the way math people use Greek letters.
112The same concept should in general get the same name across
113different contexts. When you're working in a given area, you'll
114tend to remember the binding from TLV to concept by sheer power
115of associative memory. When you come back to it, it's not that
116hard to relearn. And of course, when in doubt, comment it.
118Variables take pseudo-Hungarian suffixes, matching in general the
119suffix of the integer type:
121 c3_w wor_w; // 32-bit word
123Unlike in standard Hungarian, there is no change for pointer
124variables. C structure variables take a `_u` suffix.
128The code (from `defs.h`) tells the story:
133 # define _(x) (c3y == (x))
134 # define __(x) ((x) ? c3y : c3n)
135 # define c3a(x, y) __(_(x) && _(y))
136 # define c3o(x, y) __(_(x) || _(y))
138In short, use `_()` to turn a loobean into a boolean, `__` to go
139the other way. Use `!` as usual, `c3y` for yes and `c3n` for no,
140`c3a` for and and `c3o` for or.
144The division between `c3` and `u3` is that you could theoretically
145imagine using `c3` as just a generic C environment. Anything to do
146with nouns is in `u3`.
148### u3: a map of the system
150There are two kinds of symbols in `u3`: regular and irregular.
151Regular symbols follow this pattern:
154 -------------------------------------------------------
155 u3a_ allocation i/n/a.h n/a.c
156 u3e_ persistence i/n/e.h n/e.c
157 u3h_ hashtables i/n/h.h n/h.c
158 u3i_ noun construction i/n/i.h n/i.c
159 u3j_ jet control i/n/j.h n/j.c
160 u3m_ system management i/n/m.h n/m.c
161 u3n_ nock computation i/n/n.h n/n.c
162 u3r_ noun access (error returns) i/n/r.h n/r.c
163 u3t_ profiling i/n/t.h n/t.c
164 u3v_ arvo i/n/v.h n/v.c
165 u3x_ noun access (error crashes) i/n/x.h n/x.c
166 u3z_ memoization i/n/z.h n/z.c
167 u3k[a-g] jets (transfer, C args) i/j/k.h j/[a-g]/*.c
168 u3q[a-g] jets (retain, C args) i/j/q.h j/[a-g]/*.c
169 u3w[a-g] jets (retain, nock core) i/j/w.h j/[a-g]/*.c
171Irregular symbols always start with `u3` and obey no other rules.
172They're defined in `i/n/u.h`. Finally, `i/all.h` includes all
173these headers (fast compilers, yay) and is all you need to
176### u3: noun internals
178A noun is a `u3_noun` - currently defined as a 32-bit `c3_w`.
180If your `u3_noun` is less than `(1 << 31)`, it's a direct atom.
181Every unsigned integer between `0` and `0x7fffffff` inclusive is
184If bit `31` is set in a `u3_noun`, bit `30` is always set - this
185bit is reserved. Bit `29` is `1` if the noun is a cell, `0` if
186it's an atom. Bits `28` through `0` are a word pointer into the
187loom - see below. The structures are:
192 c3_w buf_w[0]; // actually [len_w]
201The only thing that should be mysterious here is `mug_w`, which
202is a 31-bit lazily computed nonzero short hash (FNV currently,
203soon Murmur3). If `mug_w` is 0, the hash is not yet computed.
204We also hijack this field for various hacks, such as saving the
205new address of a noun when copying over.
207Also, the value `0xffffffff` is `u3_none`, which is never a valid
208noun. Use the type `u3_weak` to express that a noun variable may
211### u3: reference counts
213The only really essential thing you need to know about `u3` is
214how to handle reference counts. Everything else, you can skip
217u3 deals with reference-counted, immutable, acyclic nouns.
218Unfortunately, we are not Apple and can't build reference
219counting into your C compiler, so you need to count by hand.
221Every allocated noun (or any allocation object, because our
222allocator is general-purpose) contains a counter which counts the
223number of references to it - typically variables with type
224`u3_noun`. When this counter goes to 0, the noun is freed.
226To tell `u3` that you've added a reference to a noun, call the
227function `u3a_gain()` or its shorthand `u3k()`. (For your
228convenience, this function returns its argument.) To tell `u3`
229that you've destroyed a reference, call `u3a_lose()` or `u3z()`.
231(If you screw up by decrementing the counter too much, `u3` will
232dump core in horrible ways. If you screw up by incrementing it
233too much, `u3` will leak memory. To check for memory leaks,
234set the `bug_o` flag in `u3e_boot()` - eg, run `vere` with `-g`.
235Memory leaks are difficult to debug - the best way to handle
236leaks is just to revert to a version that didn't have them, and
237look over your code again.)
239(You can gain or lose a direct atom. It does nothing.)
241### u3: reference protocols
243*THIS IS THE MOST CRITICAL SECTION IN THE `u3` DOCUMENTATION.*
245The key question when calling a C function in a refcounted world
246is what the function will do to the noun refcounts - and, if the
247function returns a noun, what it does to the return.
249There are two semantic patterns, `transfer` and `retain`. In
250`transfer` semantics, the caller "gives" a use count to the
251callee, which "gives back" any return. For instance, if I have
254 u3_noun foo = u3i_string("foobar");
262Suppose `u3f_futz()` has `transfer` semantics. At `[...]`, my
263code holds one reference to `bar` and zero references to `foo` -
264which has been freed, unless it's part of `bar`. My code now
265owns `bar` and gets to work with it until it's done, at which
266point a `u3z()` is required.
268On the other hand, if `u3f_futz()` has `retain` semantics, we
272 u3_noun foo = u3i_string("foobar");
280because calling `u3f_futz()` does not release our ownership of
281`foo`, which we have to free ourselves.
283But if we free `bar`, we are making a great mistake, because our
284reference to it is not in any way registered in the memory
285manager (which cannot track references in C variables, of
286course). It is normal and healthy to have these uncounted
287C references, but they must be treated with care.
289The bottom line is that it's essential for the caller to know
290the refcount semantics of any function which takes or returns a
291noun. (In some unusual circumstances, different arguments or
292returns in one function may be handled differently.)
294Broadly speaking, as a design question, retain semantics are more
295appropriate for functions which inspect or query nouns. For
296instance, `u3h()` (which takes the head of a noun) retains, so
297that we can traverse a noun tree without constantly incrementing
300Transfer semantics are more appropriate for functions which make
301nouns, which is obviously what most functions do.
303In general, though, in most places it's not worth thinking about
304what your function does. There is a convention for it, which
305depends on where it is, not what it does. Follow the convention.
307### u3: reference conventions
309The `u3` convention is that, unless otherwise specified, *all
310functions have transfer semantics* - with the exception of the
311prefixes: `u3r`, `u3x`, `u3z`, `u3q` and `u3w`. Also, within
312jet directories `a` through `f` (but not `g`), internal functions
313retain (for historical reasons).
315If functions outside this set have retain semantics, they need to
316be commented, both in the `.h` and `.c` file, with `RETAIN` in
317all caps. Yes, it's this important.
319### u3: system architecture
321If you just want to tinker with some existing code, it might be
322enough to understand the above. If not, it's probably worth
323taking the time to look at `u3` as a whole.
325`u3` is designed to work as a persistent event processor.
326Logically, it computes a function of the form
328 f(event, old state) -> (actions, new state)
330Obviously almost any computing model - including, but not limited
331to, Urbit - can be defined in this form. To create the illusion
332of a computer that never loses state and never fails, we:
334- log every event externally before it goes into u3
335- keep a single reference to a permanent state noun.
336- can abort any event without damaging the permanent state.
337- snapshot the permanent state periodically, and/or prune logs.
339### u3: the road model
341`u3` uses a memory design which I'm sure someone has invented
342somewhere before, because it's not very clever, but I've never
343seen it anywhere in particular.
345Every allocation starts with a solid block of memory, which `u3`
346calls the `loom`. How do we allocate on the loom? You're
347probably familiar with the Unix heap-stack design, in which the
348stack grows downward and the heap (malloc arena) grows upward:
352 |------------#################################+++++++++++++|
356A road is a normal heap-stack system, except that the heap
357and stack can point in *either direction*. Therefore, inside
358a road, we can nest another road in the *opposite direction*.
360When the opposite road completes, its heap is left on top of
361the opposite heap's stack. It's no more than the normal
362behavior of a stack machine for all subcomputations to push
363their results on the stack.
365The performance tradeoff of "leaping" - reversing directions in
366the road - is that if the outer computation wants to preserve the
367results of the inner one, not just use them for temporary
368purposes, it has to *copy them*.
370This is a trivial cost in some cases, a prohibitive cost in
371others. The upside, of course, is that all garbage accrued
372in the inner computation is discarded at zero cost.
374The goal of the road system is the ability to *layer* memory
375models. If you are allocating on a road, you have no idea
376how deep within a nested road system you are - in other words,
377you have no idea exactly how durable your result may be.
378But free space is never fragmented within a road.
380Roads do not reduce the generality or performance of a memory
381system, since even the most complex GC system can be nested
382within a road at no particular loss of performance - a road
383is just a block of memory.
385Each road (`u3a_road` to be exact) uses four pointers: `rut` is
386the bottom of the arena, `hat` the top of the arena, `mat` the
387bottom of the stack, `cap` the top of the stack. (Bear in mind
388that the road "stack" is not actually used as the C function-call
389stack, though it probably should be.)
391A "north" road has the stack high and the heap low:
395 |~~~~~~~~~~~~-------##########################+++++++$~~~~~|
399A "south" road is the other way around:
403 |~~~~~~~~~~~~$++++++##########################--------~~~~~|
407Legend: `-` is durable storage (heap); `+` is temporary storage
408(stack); `~` is deep storage (immutable); `$` is the allocation
409frame `#` is free memory.
411Pointer restrictions: pointers stored in `+` can point anywhere.
412Pointers in `-` can only point to `-` or `~`; pointers in `~`
415To "leap" is to create a new inner road in the `###` free space.
416but in the reverse direction, so that when the inner road
417"falls" (terminates), its durable storage is left on the
418temporary storage of the outer road.
420`u3` keeps a global variable, `u3_Road` or its alias `u3R`, which
421points to the current road. (If we ever run threads in inner
422roads - see below - this will become a thread-local variable.)
423Relative to `u3R`, `+` memory is called `junior` memory; `-`
424memory is `normal` memory; `~` is `senior` memory.
426### u3: explaining the road model
430We're now ready to understand why the road system works so
431logically with the event and persistence model.
433The key is that *we don't update refcounts in senior memory.*
434A pointer from an inner road to an outer road is not counted.
435Also, the outmost, or `surface` road, is the only part of the
436image that gets checkpointed.
438So the surface road contains the entire durable state of `u3`.
439When we process an event, or perform any kind of complicated or
440interesting calculation, *we process it in an inner road*. If
441its results are saved, they need to be copied.
443Since processing in an inner road does not touch surface memory,
444(a) we can leave the surface road in a read-only state and not
445mark its pages dirty; (b) we can abort an inner calculation
446without screwing up the surface; and (c) because inner results
447are copied onto the surface, the surface doesn't get fragmented.
449All of (a), (b) and (c) are needed for checkpointing to be easy.
450It might be tractable otherwise, but easy is even better.
452Moreover, while the surface is most definitely single-threaded,
453we could easily run multiple threads in multiple inner roads
454(as long as the threads don't have pointers into each others'
455memory, which they obviously shouldn't).
457Moreover, in future, we'll experiment more with adding road
458control hints to the programmer's toolbox. Reference counting is
459expensive. We hypothesize that in many - if not most - cases,
460the programmer can identify procedural structures whose garbage
461should be discarded in one step by copying the results. Then,
462within the procedure, we can switch the allocator into `sand`
463mode, and stop tracking references at all.
465### u3: rules for C programming
467There are two levels at which we program in C: (1) above the
468interpreter; (2) within the interpreter or jets. These have
469separate rules which need to be respected.
471### u3: rules above the interpreter
473In its relations with Unix, Urbit follows a strict rule of "call
474me, I won't call you." We do of course call Unix system calls,
475but only for the purpose of actually computing.
477Above Urbit, you are in a normal C/Unix programming environment
478and can call anything in or out of Urbit. Note that when using
479`u3`, you're always on the surface road, which is not thread-safe
480by default. Generally speaking, `u3` is designed to support
481event-oriented, single-threaded programming.
483If you need threads which create nouns, you could use
484`u3m_hate()` and `u3m_love()` to run these threads in subroads.
485You'd need to make the global road pointer, `u3R`, a thread-local
486variable instead. This seems perfectly practical, but we haven't
487done it because we haven't needed to.
489### u3: rules within the interpreter
491Within the interpreter, your code can run either in the surface
492road or in a deep road. You can test this by testing
496ie: does the pier's home road equal the current road pointer?
498Normally in this context you assume you're obeying the rules of
499running on an inner road, ie, "deep memory." Remember, however,
500that the interpreter *can* run on surface memory - but anything
501you can do deep, you can do on the surface. The converse is by
504In deep memory, think of yourself as if in a signal handler.
505Your execution context is extremely fragile and may be terminated
506without warning or cleanup at any time (for instance, by ^C).
508For instance, you can't call `malloc` (or C++ `new`) in your C
509code, because you don't have the right to modify data structures
510at the global level, and will leave them in an inconsistent state
511if your inner road gets terminated. (Instead, use our drop-in
512replacements, `u3a_malloc()`, `u3a_free()`, `u3a_realloc()`.)
514A good example is the different meaning of `c3_assert()` inside
515and outside the interpreter. At either layer, you can use
516regular assert(), which will just kill your process. On the
517surface, `c3_assert()` will just... kill your process.
519In deep execution, `c3_assert()` will issue an exception that
520queues an error event, complete with trace stack, on the Arvo
521event queue. Let's see how this happens.
525You produce an exception with
527 /* u3m_bail(): bail out. Does not return.
531 ** %exit :: semantic failure
532 ** %evil :: bad crypto
533 ** %intr :: interrupt
534 ** %fail :: execution failure
535 ** %foul :: assert failure
536 ** %need :: network block
537 ** %meme :: out of memory
538 ** %time :: timed out
539 ** %oops :: assertion failure
542 u3m_bail(c3_m how_m);
544Broadly speaking, there are two classes of exception: internal
545and external. An external exception begins in a Unix signal
546handler. An internal exception begins with a call to longjmp()
549There are also two kinds of exception: mild and severe. An
550external exception is always severe. An internal exception is
551normally mild, but some (like `c3__oops`, generated by
552`c3_assert()`) are severe.
554Either way, exceptions come with a stack trace. The `u3` nock
555interpreter is instrumented to retain stack trace hints and
556produce them as a printable `(list tank)`.
558Mild exceptions are caught by the first virtualization layer and
559returned to the caller, following the behavior of the Nock
560virtualizer `++mock` (in `hoon.hoon`)
562Severe exceptions, or mild exceptions at the surface, terminate
563the entire execution stack at any depth and send the cumulative
564trace back to the `u3` caller.
566For instance, `vere` uses this trace to construct a `%crud`
567event, which conveys our trace back toward the Arvo context where
568it crashed. This lets any UI component anywhere, even on a
569remote node, render the stacktrace as a consequence of the user's
570action - even if its its direct cause was (for instance) a Unix
573### u3: C structures on the loom
575Normally, all data on the loom is nouns. Sometimes we break this
576rule just a little, though - eg, in the `u3h` hashtables.
578To point to non-noun C structs on the loom, we use a `u3_post`,
579which is just a loom word offset. A macro lets us declare this
580as if it was a pointer:
582 typedef c3_w u3_post;
583 #define u3p(type) u3_post
585Some may regard this as clever, others as pointless. Anyway, use
586`u3to()` and `u3of()` to convert to and from pointers.
588When using C structs on the loom - generally a bad idea - make
589sure anything which could be on the surface road is structurally
590portable, eg, won't change size when the pointer size changes.
591(Note also: we consider little-endian, rightly or wrongly, to
592have won the endian wars.)
594## u3: API overview by prefix
596Let's run through the `u3` modules one by one. All public
597functions are commented, but the comments may be cryptic.
603 /* u3m_boot(): start the u3 system.
606 u3m_boot(c3_o nuu_o, c3_o bug_o, c3_c* dir_c);
608`nuu_o` is `c3y` (yes, `0`) if you're creating a new pier,
609`c3n` (no, `1`) if you're loading an existing one. `bug_o`
610is `c3y` if you want to test the garbage-collector, `c3n`
611otherwise. `dir_c` is the directory for the pier files.
613`u3m_boot()` expects an `urbit.pill` file to load the kernel
614from. It will try first `$dir/.urb.urbit.pill`, then `U3_LIB`.
616Any significant computation with nouns, certainly anything Turing
617complete, should be run (a) virtualized and (b) in an inner road.
618These are slightly different things, but at the highest level we
619bundle them together for your convenience, in `u3m_soft()`:
621 /* u3m_soft(): system soft wrapper. unifies unix and nock errors.
623 ** Produces [%$ result] or [%error (list tank)].
626 u3m_soft(c3_w sec_w, u3_funk fun_f, u3_noun arg);
628`sec_w` is the number of seconds to time out the computation.
629`fun_f` is a C function accepting `arg`.
631The result of `u3m_soft()` is a cell whose head is an atom. If
632the head is `%$` - ie, `0` - the tail is the result of
633`fun_f(arg)`. Otherwise, the head is a `term` (an atom which is
634an LSB first string), and the tail is a `(list tank)` (a list of
635`tank` printables - see `++tank` in `hoon.hoon`). Error terms
636should be the same as the exception terms above.
638If you're confident that your computation won't fail, you can
639use `u3m_soft_sure()`, `u3m_soft_slam()`, or `u3m_soft_nock()`
640for C functions, Hoon function calls, and Nock invocations.
641Caution - this returns just the result, and asserts globally.
643All the `u3m_soft` functions above work *only on the surface*.
644Within the surface, virtualize with `u3m_soft_run()`. Note that
645this takes a `fly` (a namespace gate), thus activating the `11`
646super-operator in the nock virtualizer, `++mock`. When actually
647using the `fly`, call `u3m_soft_esc()`. Don't do either unless
648you know what you're doing!
650For descending into a subroad *without* Nock virtualization,
651use `u3m_hate()` and `u3m_love` respectively. Hating enters
652a subroad; loving leaves it, copying out a product noun.
654Other miscellaneous tools in `u3m`: `u3m_file()` loads a Unix
655file as a Nock atom; `u3m_water()` measures the boundaries of
656the loom in current use (ie, watermarks); and a variety of
657prettyprinting routines, none perfect, are available, mainly for
658debugging printfs: `u3m_pretty()`, `u3m_p()`, `u3m_tape()` and
661It's sometimes nice to run a mark-and-sweep garbage collector,
662`u3m_grab()`, which collects the world from a list of roots,
663and asserts if it finds any leaks or incorrect refcounts. This
664tool is for debugging and long-term maintenance only; refcounts
669The jet system, `u3j`, is what makes `u3` and `nock` in any sense
670a useful computing environment. Except perhaps `u3a` (there is
671really no such thing as a trivial allocator, though `u3a` is
672dumber than most) - `u3j` is the most interesting code in `u3`.
674Let's consider the minor miracle of driver-to-battery binding
675which lets `u3j` work - and decrement not be `O(n)` - without
676violating the precisely defined semantics of pure Nock, *ever*.
678It's easy to assume that jets represent an architectural coupling
679between Hoon language semantics and Nock interpreter internals.
680Indeed such a coupling would be wholly wrongtious and un-Urbit.
681But the jet system is not Hoon-specific. It is specific to nock
682runtime systems that use a design pattern we call a `core`.
684#### u3j: core structure
686A core is no more than a cell `[code data]`, in which a `code` is
687either a Nock formula or a cell of `code`s, and `data` is anything.
688In a proper core, the subject each formula expects is the core
691Except for the arbitrary decision to make a core `[code data]`,
692(or as we sometimes say, `[battery payload]`), instead of `[data
693code]`, any high-level language transforming itself to Nock would
696So jets are in fact fully general. Broadly speaking, the jet
697system works by matching a C *driver* to a battery. When the
698battery is invoked with Nock operator `9`, it must be found in
699associative memory and linked to its driver. Then we link the
700formula axis of the operation (`a` in `[9 a b]`) to a specific
701function in the driver.
703To validate this jet binding, we need to know two things. One,
704we need to know the C function actually is a perfect semantic
705match for the Nock formula. This can be developed with driver
706test flags, which work, and locked down with a secure formula
707hash in the driver, which we haven't bothered with just yet.
708(You could also try to develop a formal method for verifying
709that C functions and Nock formulas are equivalent, but this is
710a research problem for the future.)
712Two, we need to validate that the payload is appropriate for the
713battery. We should note that jets are a Nock feature and have no
714reference to Hoon. A driver which relies on the Hoon type system
715to only pair it with valid payloads is a broken driver, and
716breaks the Nock compliance of the system as a whole. So don't.
718Now, a casual observer might look at `[battery payload]` and
719expect the simplest case of it to be `[formula subject]`. That
720is: to execute a simple core whose battery is a single formula,
725Then, naturally, when we go from Hoon or a high-level language
726containing functions down to Nock, `[function arguments]` turns
727into `[formula subject]`. This seems like an obvious design, and
728we mention it only because it is *completely wrong*.
730Rather, to execute a one-armed core like the above, we run
734and the normal structure of a `gate`, which is simply Urbitese
737 [formula [sample context]]
739where `sample` is Urbitese for "arguments" - and `context`, any
740Lisper will at once recognize, is Urbitese for "environment."
742To `slam` or call the gate, we simply replace the default sample
743with the caller's data, then nock the formula on the entire gate.
745What's in the context? Unlike in most dynamic languages, it is
746not some secret system-level bag of tricks. Almost always it is
747another core. This onion continues until at the bottom, there is
748an atomic constant, conventionally is the kernel version number.
750Thus a (highly desirable) `static` core is one of the form
753 [battery static-core]
755ie, a solid stack of nested libraries without any dynamic data.
756The typical gate will thus be, for example,
758 [formula [sample [battery battery battery constant]]]
760but we would be most foolish to restrict the jet mechanism to
761cores of this particular structure. We cannot constrain a
762payload to be `[sample static-core]`, or even `[sample core]`.
763Any such constraint would not be rich enough to handle Hoon,
764let alone other languages.
768There are two fundamental rules of computer science: (1) every
769system is best understood through its state; (2) less state is
770better than more state. Sadly, a pier has three different jet
771state systems: `cold`, `warm` and `hot`. It needs all of them.
773Hot state is associated with this particular Unix process. The
774persistent pier is portable not just between process and process,
775but machine and machine or OS and OS. The set of jets loaded
776into a pier may itself change (in theory, though not in the
777present implementation) during the lifetime of the process. Hot
778state is a pure C data structure.
780Cold state is associated with the logical execution history of
781the pier. It consists entirely of nouns and ignores restarts.
783Warm state contains all dependencies between cold and hot
784state. It consists of C structures allocated on the loom.
786Warm state is purely a function of cold and hot states, and
787we can wipe and regenerate it at any time. On any restart where
788the hot state might have changed, we clear the warm state
791There is only one hot state, the global jet dashboard
792`u3j_Dash` or `u3D` for short. In the present implementation,
793u3D is a static structure not modified at runtime, except for
794numbering itself on process initialization. This structure -
795which embeds function pointers to all the jets - is defined
796in `j/tree.c`. The data structures:
798 /* u3j_harm: driver arm.
800 typedef struct _u3j_harm {
801 c3_c* fcs_c; // `.axe` or name
802 u3_noun (*fun_f)(u3_noun); // compute or 0 / semitransfer
803 c3_o ice; // perfect (don't test)
804 c3_o tot; // total (never punts)
805 c3_o liv; // live (enabled)
808 /* u3j_core: C core driver.
810 typedef struct _u3j_core {
811 c3_c* cos_c; // control string
812 struct _u3j_harm* arm_u; // blank-terminated static list
813 struct _u3j_core* dev_u; // blank-terminated static list
814 struct _u3j_core* par_u; // dynamic parent pointer
815 c3_l jax_l; // dynamic jet index
818 /* u3e_dash, u3_Dash, u3D: jet dashboard singleton
820 typedef struct _u3e_dash {
821 u3j_core* dev_u; // null-terminated static list
822 c3_l len_l; // ray_u filled length
823 c3_l all_l; // ray_u allocated length
824 u3j_core* ray_u; // dynamic driver array
827Warm and cold state is *per road*. In other words, as we nest
828roads, we also nest jet state. The jet state in the road is:
830 struct { // jet dashboard
831 u3p(u3h_root) har_p; // warm state
832 u3_noun das; // cold state
835In case you understand Hoon, `das` (cold state) is a `++dash`,
836and `har_p` (warm state) is a map from battery to `++calx`:
838 ++ bane ,@tas :: battery name
839 ++ bash ,@uvH :: label hash
840 ++ bosh ,@uvH :: local battery hash
841 ++ batt ,* :: battery
843 $: jax=,@ud :: hot core index
844 hap=(map ,@ud ,@ud) :: axis/hot arm index
845 lab=path :: label as path
846 jit=* :: arbitrary data
848 ++ calx (trel calf (pair bash cope) club) :: cached by battery
849 ++ clog (pair cope (map batt club)) :: identity record
850 ++ club (pair corp (map term nock)) :: battery pattern
851 ++ cope (trel bane axis (each bash noun)) :: core pattern
853 ++ corp (each core batt) :: parent or static
854 ++ dash (map bash clog) :: jet system
856The driver index `jax` in a `++calx` is an index into `ray_u` in the
857dashboard - ie, a pointer into hot state. This is why the warm
858state has to be reset when we reload the pier in a new process.
860Why is jet state nested? Nock of course is a functional system,
861so as we compute we don't explicitly create state. Jet state is
862an exception to this principle (which works only because it can't
863be semantically detected from Nock/Hoon) - but it can't violate
864the fundamental rules of the allocation system.
866For instance, when we're on an inner road, we can't allocate on
867an outer road, or point from an outer road to an inner. So if we
868learn something - like a mapping from battery to jet - in the
869inner road, we have to keep it in the inner road.
871Mitigating this problem, when we leave an inner road (with
872`u3m_love()`), we call `u3j_reap()` to promote jet information in
873the dying road. Reaping promotes anything we've learned about
874any battery that either (a) already existed in the outer road, or
875(b) is being saved to the outer road.
879Jet binding starts with a `%fast` hint. (In Hoon, this is
880produced by the runes `~%`, for the general case, or `~/`
881for simple functions.) To bind a jet, execute a formula of the
884 [10 [%fast clue-formula] core-formula]
886`core-formula` assembles the core to be jet-propelled.
887`clue-formula` produces the hint information, or `++clue`
888above, which we want to annotate it with.
890A clue is a triple of name, parent, and hooks:
892 ++ clue (trel chum nock (list (pair term nock)))
894The name, or `++chum`, has a bunch of historical structure which
895we don't need (cleaning these things up is tricky), but just gets
896flattened into a term.
898The parent axis is a nock formula, but always reduces to a simple
899axis, which is the address of this core's *parent*. Consider
900again an ordinary gate
902 [formula [sample context]]
904Typically the `context` is itself a library core, which itself
905has a jet binding. If so, the parent axis of this gate is `7`.
907If the parent is already bound - and the parent *must* be already
908bound, in this road or a road containing it - we can hook this core
909bottom-up into a tree hierarchy. Normally the child core is
910produced by an arm of the parent core, so this is not a problem -
911we wouldn't have the child if we hadn't already made the parent.
913The clue also contains a list of *hooks*, named nock formulas on
914the core. Usually these are arms, but they need not be. The
915point is that we often want to call a core from C, in a situation
916where we have no type or other source information. A common case
917of this is a complex system in which we're mixing functions which
918are jet-propelled with functions that aren't.
920In any case, all the information in the `%fast` hint goes to
921`u3j_mine()`, which registers the battery in cold state (`das` in
922`jed` in `u3R`), then warm state (`har_p` in `jed`).
924It's essential to understand that the `%fast` hint has to be,
925well, fast - because we apply it whenever we build a core. For
926instance, if the core is a Hoon gate - a function - we will call
927`u3j_mine` every time the function is called.
929### u3j: the cold jet dashboard
931For even more fun, the jet tree is not actually a tree of
932batteries. It's a tree of battery *labels*, where a label is
933an [axis term] path from the root of the tree. (At the root,
934if the core pattern is always followed properly, is a core whose
935payload is an atomic constant, conventionally the Hoon version.)
937Under each of these labels, it's normal to have an arbitrary
938number of different Nock batteries (not just multiple copies
939of the same noun, a situation we *do* strive to avoid). For
940instance, one might be compiled with debugging hints, one not.
942We might even have changed the semantics of the battery without
943changing the label - so long as those semantics don't invalidate
946et tree. For instance, it's normal to have
947two equivalent Nock batteries at the same time in one pier: one
948battery compiled with debugging hints, one not.
950Rather, the jet tree is a semantic hierarchy. The root of the
951hierarchy is a constant, by convention the Hoon kernel version
952because any normal jet-propelled core has, at the bottom of its
953onion of libraries, the standard kernel. Thus if the core is
955 [foo-battery [bar-battery [moo-battery 164]]]
957we can reverse the nesting to construct a hierarchical core
962extends the static core `164/moo/bar` by wrapping the `foo`
963battery (ie, in Hoon, `|%`) around it. With the core above,
964you can compute `foo` stuff, `bar` stuff, and `moo` stuff.
967Not all cores are static, of course - they may contain live data,
968like the sample in a gate (ie, argument to a function). Once
969again, it's important to remember that we track jet bindings not
970by the core, which may not be static, but by the battery, which
973(And if you're wondering how we can use a deep noun like a Nock
974formula or battery as a key in a key-value table, remember
975`mug_w`, the lazily computed short hash, in all boxed nouns.)
977In any case, `das`, the dashboard, is a map from `bash` to jet
978location record (`++clog`). A `clog` in turn contains two kinds
979of information: the `++cope`, or per-location noun; and a map of
980batteries to a per-battery `++club`.
982The `cope` is a triple of `++bane` (battery name, right now just
983a `term`); `++axis`, the axis, within *this* core, of the parent;
984and `(each bash noun)`, which is either `[0 bash]` if the parent
985is another core, or `[1 noun]`, for the constant noun (like
986`164`) if there is no parent core.
988A `bash` is just the noun hash (`++sham`) of a `cope`, which
989uniquely expresses the battery's hierarchical location without
990depending on the actual formulas.
992The `club` contains a `++corp`, which we use to actually validate
993the core. Obviously jet execution has to be perfectly compatible
994with Nock. We search on the battery, but getting the battery
995right is not enough - a typical battery is dependent on its
996context. For example, your jet-propelled library function is
997very likely to call `++dec` or other advanced kernel technology.
998If you've replaced the kernel in your context with something
999else, we need to detect this and not run the jet.
1001There are two cases for a jet-propelled core - either the entire
1002core is a static constant, or it isn't. Hence the definition
1005 ++ corp (each core batt) :: parent or static
1007Ie, a `corp` is `[0 core]` or `[1 batt]`. If it's static -
1008meaning that the jet only works with one specific core, ie, the
1009parent axis of each location in the hierarchy is `3` - we can
1010validate with a single comparison. Otherwise, we have to recurse
1011upward by checking the parent.
1013Note that there is at present no way to force a jet to depend on
1016### u3j: the warm jet dashboard
1018We don't use the cold state to match jets as we call them. We
1019use the cold state to register jets as we find them, and also to
1020rebuild the warm state after the hot state is reset.
1022What we actually use at runtime is the warm state, `jed->har_p`,
1023which is a `u3h` (built-in hashtable), allocated on the loom,
1024from battery to `++calx`.
1026A `calx` is a triple of a `++calf`, a `[bash cope]` cell, and a
1027`club`. The latter two are all straight from cold state.
1029The `calf` contains warm data dependent on hot state. It's a
1030quadruple: of `jax`, the hot driver index (in `ray_u` in
1031`u3j_dash`); `hap`, a table from arm axis (ie, the axis of each
1032formula within the battery) to driver arm index (into `arm_u` in
1033`u3j_core`); `lab`, the complete label path; and `jit`, any
1034other dynamic data that may speed up execution.
1036We construct `hap`, when we create the calx, by iterating through
1037the arms registered in the `u3j_core`. Note the way a `u3j_harm`
1038declares itself, with the string `fcs_c` which can contain either
1039an axis or a name. Most jetted cores are of course gates, which
1040have one formula at one axis within the core: `fcs_c` is `".3"`.
1042But we do often have fast cores with more complex arm structure,
1043and it would be sad to have to manage their axes by hand. To use
1044an `fcs_c` with a named arm, it's sufficient to make sure the
1045name is bound to a formula `[0 axis]` in the hook table.
1047`jit`, as its name suggests, is a stub where any sort of
1048optimization data computed on battery registration might go. To
1049use it, fill in the `_cj_jit()` function.
1051### u3j: the hot dashboard
1053Now it should be easy to see how we actually invoke jets. Every
1054time we run a nock `9` instruction (pretty often, obviously), we
1055have a core and an axis. We pass these to `u3j_kick()`, which
1056will try to execute them.
1058Because nouns with a reference count of 1 are precious,
1059`u3j_kick()` has a tricky reference control definition. It
1060reserves the right to return `u3_none` in the case where there is
1061no driver, or the driver does not apply for this case; in this
1062case, it retains argument `cor`. If it succeeds, though, it
1065`u3j_kick()` searches for the battery (always the head of the
1066core, of course) in the hot dashboard. If the battery is
1067registered, it searches for the axis in `hap` in the `calx`.
1068If it exists, the core matches a driver and the driver jets this
1069arm. If not, we return `u3_none`.
1071Otherwise, we call `fun_f` in our `u3j_harm`. This obeys the
1072same protocol as `u3j_kick()`; it can refuse to function by
1073returning `u3_none`, or consume the noun.
1075Besides the actual function pointer `fun_f`, we have some flags
1076in the `u3j_harm` which tell us how to call the arm function.
1078If `ice` is yes (`&`, `0`), the jet is known to be perfect and we
1079can just trust the product of `fun_f`. Otherwise, we need to run
1080*both* the Nock arm and `fun_f`, and compare their results.
1082(Note that while executing the C side of this test, we have to
1083set `ice` to yes; on the Nock side, we have to set `liv` to no.
1084Otherwise, many non-exponential functions become exponential.
1085When auto-testing jets in this way, the principle is that the
1086test is on the outermost layer of recursion.)
1088(Note also that anyone who multi-threads this execution
1089environment has a slight locking problem with these flags if arm
1090testing is multi-threaded.)
1092If `tot` is yes, (`&`, `0`), the arm function is *total* and has
1093to return properly (though it can still return *u3_none*).
1094Otherwise, it is *partial* and can `u3_cm_bail()` out with
1095c3__punt. This feature has a cost: the jet runs in a subroad.
1097Finally, if `liv` is no (`|`, 1), the jet is off and doesn't run.
1099It should be easy to see how the tree of cores gets declared -
1100precisely, in `j/dash.c`. We declare the hierarchy as a tree
1101of `u3j_core` structures, each of which comes with a static list
1102of arms `arm_u` and sub-cores `dev_u`.
1104In `u3j_boot()`, we traverse the hierarchy, fill in parent
1105pointers `par_u`, and enumerate all `u3j_core` structures
1106into a single flat array `u3j_dash.ray_u`. Our hot state
1107then appears ready for action.
1109### u3j: jet functions
1111At present, all drivers are compiled statically into `u3`. This is
1112not a long-term permanent solution or anything. However, it will
1113always be the case with a certain amount of core functionality.
1115For instance, there are some jet functions that we need to call
1116as part of loading the Arvo kernel - like `++cue` to unpack a
1117noun from an atom. And obviously it makes sense, when jets are
1118significant enough to compile into `u3`, to export their symbols
1119in headers and the linker.
1121There are three interface prefixes for standard jet functions:
1122`u3k`, `u3q`, and `u3w`. All jets have `u3w` interfaces; most
1123have `u3q`; some have `u3k`. Of course the actual logic is
1126`u3w` interfaces use the same protocol as `fun_f` above: the
1127caller passes the entire core, which is retained if the function
1128returns `u3_none`, transferred otherwise. Why? Again, use
1129counts of 1 are special and precious for performance hackers.
1131`u3q` interfaces break the core into C arguments, *retain* noun
1132arguments, and *transfer* noun returns. `u3k` interfaces are the
1133same, except with more use of `u3_none` and other simple C
1134variations on the Hoon original, but *transfer* both arguments
1135and returns. Generally, `u3k` are most convenient for new code.
1137Following `u3k/q/w` is `[a-f]`, corresponding to the 6 logical
1138tiers of the kernel, or `g` for user-level jets. Another letter
1139is added for functions within subcores. The filename, under
1140`j/`, follows the tier and the function name.
1142For instance, `++add` is `u3wa_add(cor)`, `u3qa_add(a, b)`, or
1143`u3ka_add(a, b)`, in `j/a/add.c`. `++get` in `++by` is
1144`u3wdb_get(cor)`, `u3kdb_get(a, b)`, etc, in `j/d/by_get.c`.
1146For historical reasons, all internal jet code in `j/[a-f]`
1147*retains* noun arguments, and *transfers* noun results. Please
1148do not do this in new `g` jets! The new standard protocol is to
1149transfer both arguments and results.
1151### u3a: allocation functions
1153`u3a` allocates on the current road (u3R). Its internal
1154structures are uninteresting and typical of a naive allocator.
1156The two most-used `u3a` functions are `u3a_gain()` to add a
1157reference count, and `u3a_lose()` to release one (and free the
1158noun, if the use count is zero). For convenience, `u3a_gain()`
1159returns its argument. The pair are generally abbreviated with
1160the macros `u3k()` and `u3z()` respectively.
1162Normally we create nouns through `u3i` functions, and don't call
1163the `u3a` allocators directly. But if you do:
1165One, there are *two* sets of allocators: the word-aligned
1166allocators and the fully-aligned (ie, malloc compatible)
1167allocators. For instance, on a typical OS X setup, malloc
1168produces 16-byte aligned results - needed for some SSE
1171These allocators are *not compatible*. For 32-bit alignment
1172as used in nouns, call
1174 /* u3a_walloc(): allocate storage measured in words.
1177 u3a_walloc(c3_w len_w);
1179 /* u3a_wfree(): free storage.
1182 u3a_wfree(void* lag_v);
1184 /* u3a_wealloc(): word realloc.
1187 u3a_wealloc(void* lag_v, c3_w len_w);
1189For full alignment, call:
1191 /* u3a_malloc(): aligned storage measured in bytes.
1194 u3a_malloc(size_t len_i);
1196 /* u3a_realloc(): aligned realloc in bytes.
1199 u3a_realloc(void* lag_v, size_t len_i);
1201 /* u3a_realloc2(): gmp-shaped realloc.
1204 u3a_realloc2(void* lag_v, size_t old_i, size_t new_i);
1206 /* u3a_free(): free for aligned malloc.
1209 u3a_free(void* tox_v);
1211 /* u3a_free2(): gmp-shaped free.
1214 u3a_free2(void* tox_v, size_t siz_i);
1216There are also a set of special-purpose allocators for building
1217atoms. When building atoms, please remember that it's incorrect
1218to have a high 0 word - the word length in the atom structure
1219must be strictly correct.
1221Of course, we don't always know how large our atom will be.
1222Therefore, the standard way of building large atoms is to
1223allocate a block of raw space with `u3a_slab()`, then chop off
1224the end with `u3a_malt()` (which does the measuring itself)
1225or `u3a_mint()` in case you've measured it yourself.
1227Once again, *do not call `malloc()`* (or C++ `new`) within any
1228code that may be run within a jet. This will cause rare sporadic
1229corruption when we interrupt execution within a `malloc()`. We'd
1230just override the symbol, but `libuv` uses `malloc()` across
1231threads within its own synchronization primitives - for this to
1232work with `u3a_malloc()`, we'd have to introduce our own locks on
1233the surface-level road (which might be a viable solution).
1235### u3n: nock execution
1237The `u3n` routines execute Nock itself. On the inside, they have
1238a surprising resemblance to the spec proper (the only interesting
1239detail is how we handle tail-call elimination) and are, as one
1240would expect, quite slow. (There is no such thing as a fast tree
1243There is only one Nock, but there are lots of ways to call it.
1244(Remember that all `u3n` functions *transfer* C arguments and
1247The simplest interpreter, `u3n_nock_on(u3_noun bus, u3_noun fol)`
1248invokes Nock on `bus` (the subject) and `fol` (the formula).
1249(Why is it`[subject formula]`, not `[formula subject]`? The same
1250reason `0` is true and `1` is false.)
1252A close relative is `u3n_slam_on(u3_noun gat, u3_noun sam)`,
1253which slams a *gate* (`gat`) on a sample (`sam`). (In a normal
1254programming language which didn't talk funny and was retarded,
1255`u3n_slam_on()` would call a function on an argument.) We could
1256write it most simply as:
1259 u3n_slam_on(u3_noun gat, u3_noun sam)
1261 u3_noun pro = u3n_nock_on
1262 (u3nc(u3k(u3h(gat)),
1263 u3nc(sam, u3k(u3t(u3t(gat))))),
1269Simpler is `u3n_kick_on(u3_noun gat)`, which slams a gate (or,
1270more generally, a *trap* - because sample structure is not even
1271needed here) without changing its sample:
1274 u3n_kick_on(u3_noun gat, u3_noun sam)
1276 return u3n_nock_on(gat, u3k(u3h(gat)));
1279The `_on` functions in `u3n` are all defined as pure Nock. But
1280actually, even though we say we don't extend Nock, we do. But we
1283Note that `u3` has a well-developed error handling system -
1284`u3m_bail()` to throw an exception, `u3m_soft_*` to catch one.
1285But Nock has no exception model at all. That's okay - all it
1286means if that if an `_on` function bails, the exception is an
1287exception in the caller.
1289However, `u3`'s exception handling happens to match a convenient
1290virtual super-Nock in `hoon.hoon`, the infamous `++mock`. Of
1291course, Nock is slow, and `mock` is Nock in Nock, so it is
1292(logically) super-slow. Then again, so is decrement.
1294With the power of `u3`, we nest arbitrary layers of `mock`
1295without any particular performance cost. Moreover, we simply
1296treat Nock proper as a special case of `mock`. (More precisely,
1297the internal VM loop is `++mink` and the error compiler is
1298`++mook`. But we call the whole sandbox system `mock`.)
1300The nice thing about `mock` functions is that (by executing
1301within `u3m_soft_run()`, which as you may recall uses a nested
1302road) they provide both exceptions and the namespace operator -
1303`.^` in Hoon, which becomes operator `11` in `mock`.
1305`11` requires a namespace function, or `fly`, which produces a
1306`++unit` - `~` (`0`) for no binding, or `[0 value]`. The sample
1307to a `fly` is a `++path`, just a list of text `span`.
1309`mock` functions produce a `++toon`. Fully elaborated:
1311 ++ noun ,* :: any noun
1312 ++ path (list ,@ta) :: namespace path
1313 ++ span ,@ta :: text-atom (ASCII)
1314 ++ toon $% [%0 p=noun] :: success
1315 [%1 p=(list path)] :: blocking paths
1316 [%2 p=(list tank)] :: stack trace
1318 ++ tank :: printable
1319 $% [%leaf p=tape] :: flat text
1320 $: %palm :: backstep list
1321 p=[p=tape q=tape r=tape s=tape] :: mid cap open close
1322 q=(list tank) :: contents
1324 $: %rose :: straight list
1325 p=[p=tape q=tape r=tape] :: mid open close
1326 q=(list tank) :: contents
1330(Note that `tank` is overdesigned and due for replacement.)
1332What does a `toon` mean? Either your computation succeded (`[0
1333noun]`, or could not finish because it blocked on one or more
1334global paths (`[1 (list path)]`), or it exited with a stack trace
1337Note that of all the `u3` exceptions, only `%exit` is produced
1338deterministically by the Nock definition. Therefore, only
1339`%exit` produces a `2` result. Any other argument to
1340`u3m_bail()` will unwind the virtualization stack all the way to
1341the top - or to be more exact, to `u3m_soft_top()`.
1343In any case, the simplest `mock` functions are `u3n_nock_un()`
1344and `u3n_slam_un()`. These provide exception control without
1345any namespace change, as you can see by the code:
1347 /* u3n_nock_un(): produce .*(bus fol), as ++toon.
1350 u3n_nock_un(u3_noun bus, u3_noun fol)
1352 u3_noun fly = u3nt(u3nt(11, 0, 6), 0, 0); // |=(a=* .^(a))
1354 return u3n_nock_in(fly, bus, fol);
1357 /* u3n_slam_un(): produce (gat sam), as ++toon.
1360 u3n_slam_un(u3_noun gat, u3_noun sam)
1362 u3_noun fly = u3nt(u3nt(11, 0, 6), 0, 0); // |=(a=* .^(a))
1364 return u3n_slam_in(fly, gat, sam);
1367The `fly` is added as the first argument to `u3n_nock_in()` and
1368`u3n_slam_in()`. Of course, logically, `fly` executes in the
1369caller's exception layer. (Maintaining this illusion is slightly
1370nontrivial.) Finally, `u3n_nock_an()` is a sandbox with a null
1375The only `u3e` function you should need to call is `u3e_save()`,
1376which saves the loom. As it can be restored on any platform,
1377please make sure you don't have any state in the loom that is
1378bound to your process or architecture - except for exceptions
1379like the warm jet state, which is actively purged on reboot.
1381### u3r: reading nouns (weak)
1383As befits accessors they don't make anything, `u3r` noun reading
1384functions always retain their arguments and their returns. They
1385never bail; rather, when they don't work, they return a `u3_weak`
1388Most of these functions are straightforward and do only what
1389their comments say. A few are interesting enough to discuss.
1391`u3r_at()` is the familiar tree fragment function, `/` from the
1392Nock spec. For taking complex nouns apart, `u3r_mean()` is a
1393relatively funky way of deconstructing nouns with a varargs list
1394of `axis`, `u3_noun *`. For cells, triples, etc, decompose with
1395`u3r_cell()`, `u3r_trel()`, etc. For the tagged equivalents, use
1396`u3r_pq()` and friends.
1398`u3r_sing(u3_noun a, u3_noun b)` (true if `a` and `b` are a
1399*single* noun) are interesting because it uses mugs to help it
1400out. Clearly, different nouns may have the same mug, but the
1401same nouns cannot have a different mug. It's important to
1402understand the performance characteristics of `u3r_sing()`:
1403the worst possible case is a comparison of duplicate nouns,
1404which have the same value but were created separately. In this
1405case, the tree is traversed
1407`u3r_sung()` is a deeply funky and frightening version of
1408`u3r_sing()` that unifies pointers to the duplicate nouns it
1409finds, freeing the second copy. Obviously, do not use
1410`u3r_sung()` when you have live, but not reference counted, noun
1411references from C - if they match a noun with a refcount of 1
1412that gets freed, bad things happen.
1414It's important to remember that `u3r_mug()`, which produces a
141531-bit, nonzero insecure hash, uses the `mug_w` slot in any boxed
1416noun as a lazy cache. There are a number of variants of
1417`u3r_mug()` that can get you out of building unneeded nouns.
1419### u3x: reading nouns (bail)
1421`u3x` functions are like `u3r` functions, but instead of
1422returning `u3_none` when (for instance) we try to take the head
1423of an atom, they bail with `%exit`. In other words, they do what
1424the same operation would do in Nock.
1426### u3h: hash tables.
1428We can of course use the Hoon `map` structure as an associative
1429array. This is a balanced treap and reasonably fast. However,
1430it's considerably inferior to a custom structure like an HAMT
1431(hash array-mapped trie). We use `u3_post` to allocate HAMT
1432structures on the loom.
1434(Our HAMT implements the classic Bagwell algorithm which depends
1435on the `gcc` standard directive `__builtin_popcount()`. On a CPU
1436which doesn't support popcount or an equivalent instruction, some
1437other design would probably be preferable.)
1439There's no particular rocket science in the API. `u3h_new()`
1440creates a hashtable; `u3h_free()` destroys one; `u3h_put()`
1441inserts, `u3h_get()` retrieves. You can transform values in a
1442hashtable with `u3h_walk()`.
1444The only funky function is `u3h_gut()`, which unifies keys with
1445`u3r_sung()`. As with all cases of `u3r_sung()`, this must be
1446used with extreme caution.
1450Connected to the `~+` rune in Hoon, via the Nock `%memo` hint,
1451the memoization facility is a general-purpose cache.
1453(It's also used for partial memoization - a feature that'll
1454probably be removed, in which conservative worklist algorithms
1455(which would otherwise be exponential) memoize everything in the
1456subject *except* the worklist. This is used heavily in the Hoon
1457compiler jets (j/f/*.c). Unfortunately, it's probably not
1458possible to make this work perfectly in that it can't be abused
1459to violate Nock, so we'll probably remove it at a later date,
1460instead making `++ut` keep its own monadic cache.)
1462Each `u3z` function comes with a `c3_m` mote which disambiguates
1463the function mapping key to value. For Nock itself, use 0. For
1464extra speed, small tuples are split out in C; thus, find with
1466 u3_weak u3z_find(c3_m, u3_noun);
1467 u3_weak u3z_find_2(c3_m, u3_noun, u3_noun);
1468 u3_weak u3z_find_3(c3_m, u3_noun, u3_noun, u3_noun);
1469 u3_weak u3z_find_4(c3_m, u3_noun, u3_noun, u3_noun, u3_noun);
1473 u3_noun u3z_save(c3_m, u3_noun, u3_noun);
1474 u3_noun u3z_save_2(c3_m, u3_noun, u3_noun, u3_noun);
1475 u3_noun u3z_save_3(c3_m, u3_noun, u3_noun, u3_noun, u3_noun);
1476 u3_noun u3z_save_4(c3_m, u3_noun, u3_noun, u3_noun, u3_noun, u3_noun);
1478where the value is the last argument. To eliminate duplicate
1484`u3z` functions retain keys and transfer values.
1486The `u3z` cache, built on `u3h` hashes, is part of the current
1487road, and goes away when it goes away. (In future, we may wish
1488to promote keys/values which outlive the road, as we do with jet
1489state.) There is no cache reclamation at present, so be careful.
1491### u3t: tracing and profiling.
1495### u3v: the Arvo kernel
1497An Arvo kernel - or at least, a core that compiles with the Arvo
1498interface - is part of the global `u3` state. What is an Arvo
1499core? Slightly pseudocoded:
1503 ++ come |= [yen=@ ova=(list ovum) nyf=pone] :: 11
1504 ^- [(list ovum) _+>]
1506 ++ keep |= [now=@da hap=path] :: 4
1509 ++ load |= [yen=@ ova=(list ovum) nyf=pane] :: 86
1510 ^- [(list ovum) _+>]
1512 ++ peek |= [now=@da path] :: 87
1515 ++ poke |= [now=@da ovo=ovum] :: 42
1516 ^- [(list ovum) _+>]
1518 ++ wish |= txt=@ta :: 20
1522 ++ card ,[p=@tas q=*] :: typeless card
1523 ++ ovum ,[p=wire q=card] :: Arvo event
1524 ++ wire path :: event cause
1526This is the Arvo ABI in a very real sense. Arvo is a core with
1527these six arms. To use these arms, we hardcode the axis of the
1528formula (`11`, `4`, `86`, etc) into the C code that calls Arvo,
1529because otherwise we'd need type metadata - which we can get, by
1532It's important to understand the Arvo event/action structure, or
1533`++ovum`. An `ovum` is a `card`, which is any `[term noun]`
1534cell, and a `++wire`, a `path` which indicates the location of
1535the event. At the Unix level, the `wire` corresponds to a system
1536module or context. For input events, this is the module that
1537caused the event; for output actions, it's the module that
1540`++poke` sends Arvo an event `ovum`, producing a cell of action
1541ova and a new Arvo core.
1543`++peek` dereferences the Arvo namespace. It takes a date and a
1544key, and produces `~` (`0`) or `[~ value]`.
1546`++keep` asks Arvo the next time it wants to be woken up, for the
1547given `wire`. (This input will probably be eliminated in favor
1548of a single global timer.)
1550`++wish` compiles a string of Hoon source. While just a
1551convenience, it's a very convenient convenience.
1553`++come` and `++load` are used by Arvo to reset itself (more
1554precisely, to shift the Arvo state from an old kernel to a new
1555one); there is no need to call them from C.
1557Now that we understand the Arvo kernel interface, let's look at
1558the `u3v` API. As usual, all the functions in `u3v` are
1559commented, but unfortunately it's hard to describe this API as
1560clean at present. The problem is that `u3v` remains design
1561coupled to the old `vere` event handling code written for `u2`.
1562But let's describe the functions you should be calling, assuming
1563you're not writing the next event system. There are only two.
1565`u3v_wish(str_c)` wraps the `++wish` functionality in a cache
1566(which is read-only unless you're on the surface road).
1568`u3v_do()` uses `wish` to provide a convenient interface for
1569calling Hoon kernel functions by name. Even more conveniently,
1570we tend to call `u3v_do()` with these convenient aliases:
1572 #define u3do(txt_c, arg) u3v_do(txt_c, arg)
1573 #define u3dc(txt_c, a, b) u3v_do(txt_c, u3nc(a, b))
1574 #define u3dt(txt_c, a, b, c) u3v_do(txt_c, u3nt(a, b, c))
1575 #define u3dq(txt_c, a, b, c, d) u3v_do(txt_c, u3nt(a, b, c, d))
keysesc to close
- j k
- move down, up
- enter
- open
- esc
- back, or up a level
- 1 2 3 4
- readme, files, commits, refs
- /
- filter
- y
- copy the clone command
- c
- compare the chosen commit with another (or shift-click both)
- ?
- show or hide this list