aboutsummaryrefslogtreecommitdiff
path: root/doc/ext_buffer.html
diff options
context:
space:
mode:
Diffstat (limited to 'doc/ext_buffer.html')
-rw-r--r--doc/ext_buffer.html689
1 files changed, 689 insertions, 0 deletions
diff --git a/doc/ext_buffer.html b/doc/ext_buffer.html
new file mode 100644
index 00000000..4680f2cc
--- /dev/null
+++ b/doc/ext_buffer.html
@@ -0,0 +1,689 @@
1<!DOCTYPE html>
2<html>
3<head>
4<title>String Buffer Library</title>
5<meta charset="utf-8">
6<meta name="Copyright" content="Copyright (C) 2005-2026">
7<meta name="Language" content="en">
8<link rel="stylesheet" type="text/css" href="bluequad.css" media="screen">
9<link rel="stylesheet" type="text/css" href="bluequad-print.css" media="print">
10<style type="text/css">
11.lib {
12 vertical-align: middle;
13 margin-left: 5px;
14 padding: 0 5px;
15 font-size: 60%;
16 border-radius: 5px;
17 background: #c5d5ff;
18 color: #000;
19}
20</style>
21</head>
22<body>
23<div id="site">
24<a href="https://luajit.org"><span>Lua<span id="logo">JIT</span></span></a>
25</div>
26<div id="head">
27<h1>String Buffer Library</h1>
28</div>
29<div id="nav">
30<ul><li>
31<a href="luajit.html">LuaJIT</a>
32<ul><li>
33<a href="https://luajit.org/download.html">Download <span class="ext">&raquo;</span></a>
34</li><li>
35<a href="install.html">Installation</a>
36</li><li>
37<a href="running.html">Running</a>
38</li></ul>
39</li><li>
40<a href="extensions.html">Extensions</a>
41<ul><li>
42<a href="ext_ffi.html">FFI Library</a>
43<ul><li>
44<a href="ext_ffi_tutorial.html">FFI Tutorial</a>
45</li><li>
46<a href="ext_ffi_api.html">ffi.* API</a>
47</li><li>
48<a href="ext_ffi_semantics.html">FFI Semantics</a>
49</li></ul>
50</li><li>
51<a class="current" href="ext_buffer.html">String Buffers</a>
52</li><li>
53<a href="ext_jit.html">jit.* Library</a>
54</li><li>
55<a href="ext_c_api.html">Lua/C API</a>
56</li><li>
57<a href="ext_profiler.html">Profiler</a>
58</li></ul>
59</li><li>
60<a href="https://luajit.org/status.html">Status <span class="ext">&raquo;</span></a>
61</li><li>
62<a href="https://luajit.org/faq.html">FAQ <span class="ext">&raquo;</span></a>
63</li><li>
64<a href="https://luajit.org/list.html">Mailing List <span class="ext">&raquo;</span></a>
65</li></ul>
66</div>
67<div id="main">
68<p>
69The string buffer library allows <b>high-performance manipulation of
70string-like data</b>.
71</p>
72<p>
73Unlike Lua strings, which are constants, string buffers are
74<b>mutable</b> sequences of 8-bit (binary-transparent) characters. Data
75can be stored, formatted and encoded into a string buffer and later
76converted, extracted or decoded.
77</p>
78<p>
79The convenient string buffer API simplifies common string manipulation
80tasks, that would otherwise require creating many intermediate strings.
81String buffers improve performance by eliminating redundant memory
82copies, object creation, string interning and garbage collection
83overhead. In conjunction with the FFI library, they allow zero-copy
84operations.
85</p>
86<p>
87The string buffer library also includes a high-performance
88<a href="#serialize">serializer</a> for Lua objects.
89</p>
90
91<h2 id="use">Using the String Buffer Library</h2>
92<p>
93The string buffer library is built into LuaJIT by default, but it's not
94loaded by default. Add this to the start of every Lua file that needs
95one of its functions:
96</p>
97<pre class="code">
98local buffer = require("string.buffer")
99</pre>
100<p>
101The convention for the syntax shown on this page is that <tt>buffer</tt>
102refers to the buffer library and <tt>buf</tt> refers to an individual
103buffer object.
104</p>
105<p>
106Please note the difference between a Lua function call, e.g.
107<tt>buffer.new()</tt> (with a dot) and a Lua method call, e.g.
108<tt>buf:reset()</tt> (with a colon).
109</p>
110
111<h3 id="buffer_object">Buffer Objects</h3>
112<p>
113A buffer object is a garbage-collected Lua object. After creation with
114<tt>buffer.new()</tt>, it can (and should) be reused for many operations.
115When the last reference to a buffer object is gone, it will eventually
116be freed by the garbage collector, along with the allocated buffer
117space.
118</p>
119<p>
120Buffers operate like a FIFO (first-in first-out) data structure. Data
121can be appended (written) to the end of the buffer and consumed (read)
122from the front of the buffer. These operations may be freely mixed.
123</p>
124<p>
125The buffer space that holds the characters is managed automatically
126&mdash; it grows as needed and already consumed space is recycled. Use
127<tt>buffer.new(size)</tt> and <tt>buf:free()</tt>, if you need more
128control.
129</p>
130<p>
131The maximum size of a single buffer is the same as the maximum size of a
132Lua string, which is slightly below two gigabytes. For huge data sizes,
133neither strings nor buffers are the right data structure &mdash; use the
134FFI library to directly map memory or files up to the virtual memory
135limit of your OS.
136</p>
137
138<h3 id="buffer_overview">Buffer Method Overview</h3>
139<ul>
140<li>
141The <tt>buf:put*()</tt>-like methods append (write) characters to the
142end of the buffer.
143</li>
144<li>
145The <tt>buf:get*()</tt>-like methods consume (read) characters from the
146front of the buffer.
147</li>
148<li>
149Other methods, like <tt>buf:tostring()</tt> only read the buffer
150contents, but don't change the buffer.
151</li>
152<li>
153The <tt>buf:set()</tt> method allows zero-copy consumption of a string
154or an FFI cdata object as a buffer.
155</li>
156<li>
157The FFI-specific methods allow zero-copy read/write-style operations or
158modifying the buffer contents in-place. Please check the
159<a href="#ffi_caveats">FFI caveats</a> below, too.
160</li>
161<li>
162Methods that don't need to return anything specific, return the buffer
163object itself as a convenience. This allows method chaining, e.g.:
164<tt>buf:reset():encode(obj)</tt> or <tt>buf:skip(len):get()</tt>
165</li>
166</ul>
167
168<h2 id="create">Buffer Creation and Management</h2>
169
170<h3 id="buffer_new"><tt>local buf = buffer.new([size [,options]])<br>
171local buf = buffer.new([options])</tt></h3>
172<p>
173Creates a new buffer object.
174</p>
175<p>
176The optional <tt>size</tt> argument ensures a minimum initial buffer
177size. This is strictly an optimization when the required buffer size is
178known beforehand. The buffer space will grow as needed, in any case.
179</p>
180<p>
181The optional table <tt>options</tt> sets various
182<a href="#serialize_options">serialization options</a>.
183</p>
184
185<h3 id="buffer_reset"><tt>buf = buf:reset()</tt></h3>
186<p>
187Reset (empty) the buffer. The allocated buffer space is not freed and
188may be reused.
189</p>
190
191<h3 id="buffer_free"><tt>buf = buf:free()</tt></h3>
192<p>
193The buffer space of the buffer object is freed. The object itself
194remains intact, empty and may be reused.
195</p>
196<p>
197Note: you normally don't need to use this method. The garbage collector
198automatically frees the buffer space, when the buffer object is
199collected. Use this method, if you need to free the associated memory
200immediately.
201</p>
202
203<h2 id="write">Buffer Writers</h2>
204
205<h3 id="buffer_put"><tt>buf = buf:put([str|num|obj] [,…])</tt></h3>
206<p>
207Appends a string <tt>str</tt>, a number <tt>num</tt> or any object
208<tt>obj</tt> with a <tt>__tostring</tt> metamethod to the buffer.
209Multiple arguments are appended in the given order.
210</p>
211<p>
212Appending a buffer to a buffer is possible and short-circuited
213internally. But it still involves a copy. Better combine the buffer
214writes to use a single buffer.
215</p>
216
217<h3 id="buffer_putf"><tt>buf = buf:putf(format, …)</tt></h3>
218<p>
219Appends the formatted arguments to the buffer. The <tt>format</tt>
220string supports the same options as <tt>string.format()</tt>.
221</p>
222
223<h3 id="buffer_putcdata"><tt>buf = buf:putcdata(cdata, len)</tt><span class="lib">FFI</span></h3>
224<p>
225Appends the given <tt>len</tt> number of bytes from the memory pointed
226to by the FFI <tt>cdata</tt> object to the buffer. The object needs to
227be convertible to a (constant) pointer.
228</p>
229
230<h3 id="buffer_set"><tt>buf = buf:set(str)<br>
231buf = buf:set(cdata, len)</tt><span class="lib">FFI</span></h3>
232<p>
233This method allows zero-copy consumption of a string or an FFI cdata
234object as a buffer. It stores a reference to the passed string
235<tt>str</tt> or the FFI <tt>cdata</tt> object in the buffer. Any buffer
236space originally allocated is freed. This is <i>not</i> an append
237operation, unlike the <tt>buf:put*()</tt> methods.
238</p>
239<p>
240After calling this method, the buffer behaves as if
241<tt>buf:free():put(str)</tt> or <tt>buf:free():put(cdata,&nbsp;len)</tt>
242had been called. However, the data is only referenced and not copied, as
243long as the buffer is only consumed.
244</p>
245<p>
246In case the buffer is written to later on, the referenced data is copied
247and the object reference is removed (copy-on-write semantics).
248</p>
249<p>
250The stored reference is an anchor for the garbage collector and keeps the
251originally passed string or FFI cdata object alive.
252</p>
253
254<h3 id="buffer_reserve"><tt>ptr, len = buf:reserve(size)</tt><span class="lib">FFI</span><br>
255<tt>buf = buf:commit(used)</tt><span class="lib">FFI</span></h3>
256<p>
257The <tt>reserve</tt> method reserves at least <tt>size</tt> bytes of
258write space in the buffer. It returns an <tt>uint8_t&nbsp;*</tt> FFI
259cdata pointer <tt>ptr</tt> that points to this space.
260</p>
261<p>
262The available length in bytes is returned in <tt>len</tt>. This is at
263least <tt>size</tt> bytes, but may be more to facilitate efficient
264buffer growth. You can either make use of the additional space or ignore
265<tt>len</tt> and only use <tt>size</tt> bytes.
266</p>
267<p>
268The <tt>commit</tt> method appends the <tt>used</tt> bytes of the
269previously returned write space to the buffer data.
270</p>
271<p>
272This pair of methods allows zero-copy use of C read-style APIs:
273</p>
274<pre class="code">
275local MIN_SIZE = 65536
276repeat
277 local ptr, len = buf:reserve(MIN_SIZE)
278 local n = C.read(fd, ptr, len)
279 if n == 0 then break end -- EOF.
280 if n &lt; 0 then error("read error") end
281 buf:commit(n)
282until false
283</pre>
284<p>
285The reserved write space is <i>not</i> initialized. At least the
286<tt>used</tt> bytes <b>must</b> be written to before calling the
287<tt>commit</tt> method. There's no need to call the <tt>commit</tt>
288method, if nothing is added to the buffer (e.g. on error).
289</p>
290
291<h2 id="read">Buffer Readers</h2>
292
293<h3 id="buffer_length"><tt>len = #buf</tt></h3>
294<p>
295Returns the current length of the buffer data in bytes.
296</p>
297
298<h3 id="buffer_concat"><tt>res = str|num|buf .. str|num|buf […]</tt></h3>
299<p>
300The Lua concatenation operator <tt>..</tt> also accepts buffers, just
301like strings or numbers. It always returns a string and not a buffer.
302</p>
303<p>
304Note that although this is supported for convenience, this thwarts one
305of the main reasons to use buffers, which is to avoid string
306allocations. Rewrite it with <tt>buf:put()</tt> and <tt>buf:get()</tt>.
307</p>
308<p>
309Mixing this with unrelated objects that have a <tt>__concat</tt>
310metamethod may not work, since these probably only expect strings.
311</p>
312
313<h3 id="buffer_skip"><tt>buf = buf:skip(len)</tt></h3>
314<p>
315Skips (consumes) <tt>len</tt> bytes from the buffer up to the current
316length of the buffer data.
317</p>
318
319<h3 id="buffer_get"><tt>str, … = buf:get([len|nil] [,…])</tt></h3>
320<p>
321Consumes the buffer data and returns one or more strings. If called
322without arguments, the whole buffer data is consumed. If called with a
323number, up to <tt>len</tt> bytes are consumed. A <tt>nil</tt> argument
324consumes the remaining buffer space (this only makes sense as the last
325argument). Multiple arguments consume the buffer data in the given
326order.
327</p>
328<p>
329Note: a zero length or no remaining buffer data returns an empty string
330and not <tt>nil</tt>.
331</p>
332
333<h3 id="buffer_tostring"><tt>str = buf:tostring()<br>
334str = tostring(buf)</tt></h3>
335<p>
336Creates a string from the buffer data, but doesn't consume it. The
337buffer remains unchanged.
338</p>
339<p>
340Buffer objects also define a <tt>__tostring</tt> metamethod. This means
341buffers can be passed to the global <tt>tostring()</tt> function and
342many other functions that accept this in place of strings. The important
343internal uses in functions like <tt>io.write()</tt> are short-circuited
344to avoid the creation of an intermediate string object.
345</p>
346
347<h3 id="buffer_ref"><tt>ptr, len = buf:ref()</tt><span class="lib">FFI</span></h3>
348<p>
349Returns an <tt>uint8_t&nbsp;*</tt> FFI cdata pointer <tt>ptr</tt> that
350points to the buffer data. The length of the buffer data in bytes is
351returned in <tt>len</tt>.
352</p>
353<p>
354The returned pointer can be directly passed to C functions that expect a
355buffer and a length. You can also do bytewise reads
356(<tt>local&nbsp;x&nbsp;=&nbsp;ptr[i]</tt>) or writes
357(<tt>ptr[i]&nbsp;=&nbsp;0x40</tt>) of the buffer data.
358</p>
359<p>
360In conjunction with the <tt>skip</tt> method, this allows zero-copy use
361of C write-style APIs:
362</p>
363<pre class="code">
364repeat
365 local ptr, len = buf:ref()
366 if len == 0 then break end
367 local n = C.write(fd, ptr, len)
368 if n &lt; 0 then error("write error") end
369 buf:skip(n)
370until n >= len
371</pre>
372<p>
373Unlike Lua strings, buffer data is <i>not</i> implicitly
374zero-terminated. It's not safe to pass <tt>ptr</tt> to C functions that
375expect zero-terminated strings. If you're not using <tt>len</tt>, then
376you're doing something wrong.
377</p>
378
379<h2 id="serialize">Serialization of Lua Objects</h2>
380<p>
381The following functions and methods allow <b>high-speed serialization</b>
382(encoding) of a Lua object into a string and decoding it back to a Lua
383object. This allows convenient storage and transport of <b>structured
384data</b>.
385</p>
386<p>
387The encoded data is in an <a href="#serialize_format">internal binary
388format</a>. The data can be stored in files, binary-transparent
389databases or transmitted to other LuaJIT instances across threads,
390processes or networks.
391</p>
392<p>
393Encoding speed can reach up to 1 Gigabyte/second on a modern desktop- or
394server-class system, even when serializing many small objects. Decoding
395speed is mostly constrained by object creation cost.
396</p>
397<p>
398The serializer handles most Lua types, common FFI number types and
399nested structures. Functions, thread objects, other FFI cdata and full
400userdata cannot be serialized (yet).
401</p>
402<p>
403The encoder serializes nested structures as trees. Multiple references
404to a single object will be stored separately and create distinct objects
405after decoding. Circular references cause an error.
406</p>
407
408<h3 id="serialize_methods">Serialization Functions and Methods</h3>
409
410<h3 id="buffer_encode"><tt>str = buffer.encode(obj)<br>
411buf = buf:encode(obj)</tt></h3>
412<p>
413Serializes (encodes) the Lua object <tt>obj</tt>. The stand-alone
414function returns a string <tt>str</tt>. The buffer method appends the
415encoding to the buffer.
416</p>
417<p>
418<tt>obj</tt> can be any of the supported Lua types &mdash; it doesn't
419need to be a Lua table.
420</p>
421<p>
422This function may throw an error when attempting to serialize
423unsupported object types, circular references or deeply nested tables.
424</p>
425
426<h3 id="buffer_decode"><tt>obj = buffer.decode(str)<br>
427obj = buf:decode()</tt></h3>
428<p>
429The stand-alone function deserializes (decodes) the string
430<tt>str</tt>, the buffer method deserializes one object from the
431buffer. Both return a Lua object <tt>obj</tt>.
432</p>
433<p>
434The returned object may be any of the supported Lua types &mdash;
435even <tt>nil</tt>.
436</p>
437<p>
438This function may throw an error when fed with malformed or incomplete
439encoded data. The stand-alone function throws when there's left-over
440data after decoding a single top-level object. The buffer method leaves
441any left-over data in the buffer.
442</p>
443<p>
444Attempting to deserialize an FFI type will throw an error, if the FFI
445library is not built-in or has not been loaded, yet.
446</p>
447
448<h3 id="serialize_options">Serialization Options</h3>
449<p>
450The <tt>options</tt> table passed to <tt>buffer.new()</tt> may contain
451the following members (all optional):
452</p>
453<ul>
454<li>
455<tt>dict</tt> is a Lua table holding a <b>dictionary of strings</b> that
456commonly occur as table keys of objects you are serializing. These keys
457are compactly encoded as indexes during serialization. A well-chosen
458dictionary saves space and improves serialization performance.
459</li>
460<li>
461<tt>metatable</tt> is a Lua table holding a <b>dictionary of metatables</b>
462for the table objects you are serializing.
463</li>
464</ul>
465<p>
466<tt>dict</tt> needs to be an array of strings and <tt>metatable</tt> needs
467to be an array of tables. Both starting at index 1 and without holes (no
468<tt>nil</tt> in between). The tables are anchored in the buffer object and
469internally modified into a two-way index (don't do this yourself, just pass
470a plain array). The tables must not be modified after they have been passed
471to <tt>buffer.new()</tt>.
472</p>
473<p>
474The <tt>dict</tt> and <tt>metatable</tt> tables used by the encoder and
475decoder must be the same. Put the most common entries at the front. Extend
476at the end to ensure backwards-compatibility &mdash; older encodings can
477then still be read. You may also set some indexes to <tt>false</tt> to
478explicitly drop backwards-compatibility. Old encodings that use these
479indexes will throw an error when decoded.
480</p>
481<p>
482Metatables that are not found in the <tt>metatable</tt> dictionary are
483ignored when encoding. Decoding returns a table with a <tt>nil</tt>
484metatable.
485</p>
486<p>
487Note: parsing and preparation of the options table is somewhat
488expensive. Create a buffer object only once and recycle it for multiple
489uses. Avoid mixing encoder and decoder buffers, since the
490<tt>buf:set()</tt> method frees the already allocated buffer space:
491</p>
492<pre class="code">
493local options = {
494 dict = { "commonly", "used", "string", "keys" },
495}
496local buf_enc = buffer.new(options)
497local buf_dec = buffer.new(options)
498
499local function encode(obj)
500 return buf_enc:reset():encode(obj):get()
501end
502
503local function decode(str)
504 return buf_dec:set(str):decode()
505end
506</pre>
507
508<h3 id="serialize_stream">Streaming Serialization</h3>
509<p>
510In some contexts, it's desirable to do piecewise serialization of large
511datasets, also known as <i>streaming</i>.
512</p>
513<p>
514This serialization format can be safely concatenated and supports streaming.
515Multiple encodings can simply be appended to a buffer and later decoded
516individually:
517</p>
518<pre class="code">
519local buf = buffer.new()
520buf:encode(obj1)
521buf:encode(obj2)
522local copy1 = buf:decode()
523local copy2 = buf:decode()
524</pre>
525<p>
526Here's how to iterate over a stream:
527</p>
528<pre class="code">
529while #buf ~= 0 do
530 local obj = buf:decode()
531 -- Do something with obj.
532end
533</pre>
534<p>
535Since the serialization format doesn't prepend a length to its encoding,
536network applications may need to transmit the length, too.
537</p>
538
539<h3 id="serialize_format">Serialization Format Specification</h3>
540<p>
541This serialization format is designed for <b>internal use</b> by LuaJIT
542applications. Serialized data is upwards-compatible and portable across
543all supported LuaJIT platforms.
544</p>
545<p>
546It's an <b>8-bit binary format</b> and not human-readable. It uses e.g.
547embedded zeroes and stores embedded Lua string objects unmodified, which
548are 8-bit-clean, too. Encoded data can be safely concatenated for
549streaming and later decoded one top-level object at a time.
550</p>
551<p>
552The encoding is reasonably compact, but tuned for maximum performance,
553not for minimum space usage. It compresses well with any of the common
554byte-oriented data compression algorithms.
555</p>
556<p>
557Although documented here for reference, this format is explicitly
558<b>not</b> intended to be a 'public standard' for structured data
559interchange across computer languages (like JSON or MessagePack). Please
560do not use it as such.
561</p>
562<p>
563The specification is given below as a context-free grammar with a
564top-level <tt>object</tt> as the starting point. Alternatives are
565separated by the <tt>|</tt> symbol and <tt>*</tt> indicates repeats.
566Grouping is implicit or indicated by <tt>{…}</tt>. Terminals are
567either plain hex numbers, encoded as bytes, or have a <tt>.format</tt>
568suffix.
569</p>
570<pre>
571object → nil | false | true
572 | null | lightud32 | lightud64
573 | int | num | tab | tab_mt
574 | int64 | uint64 | complex
575 | string
576
577nil → 0x00
578false → 0x01
579true → 0x02
580
581null → 0x03 // NULL lightuserdata
582lightud32 → 0x04 data.I // 32 bit lightuserdata
583lightud64 → 0x05 data.L // 64 bit lightuserdata
584
585int → 0x06 int.I // int32_t
586num → 0x07 double.L
587
588tab → 0x08 // Empty table
589 | 0x09 h.U h*{object object} // Key/value hash
590 | 0x0a a.U a*object // 0-based array
591 | 0x0b a.U h.U a*object h*{object object} // Mixed
592 | 0x0c a.U (a-1)*object // 1-based array
593 | 0x0d a.U h.U (a-1)*object h*{object object} // Mixed
594tab_mt → 0x0e (index-1).U tab // Metatable dict entry
595
596int64 → 0x10 int.L // FFI int64_t
597uint64 → 0x11 uint.L // FFI uint64_t
598complex → 0x12 re.L im.L // FFI complex
599
600string → (0x20+len).U len*char.B
601 | 0x0f (index-1).U // String dict entry
602
603.B = 8 bit
604.I = 32 bit little-endian
605.L = 64 bit little-endian
606.U = prefix-encoded 32 bit unsigned number n:
607 0x00..0xdf → n.B
608 0xe0..0x1fdf → (0xe0|(((n-0xe0)>>8)&0x1f)).B ((n-0xe0)&0xff).B
609 0x1fe0.. → 0xff n.I
610</pre>
611
612<h2 id="error">Error handling</h2>
613<p>
614Many of the buffer methods can throw an error. Out-of-memory or usage
615errors are best caught with an outer wrapper for larger parts of code.
616There's not much one can do after that, anyway.
617</p>
618<p>
619OTOH, you may want to catch some errors individually. Buffer methods need
620to receive the buffer object as the first argument. The Lua colon-syntax
621<tt>obj:method()</tt> does that implicitly. But to wrap a method with
622<tt>pcall()</tt>, the arguments need to be passed like this:
623</p>
624<pre class="code">
625local ok, err = pcall(buf.encode, buf, obj)
626if not ok then
627 -- Handle error in err.
628end
629</pre>
630
631<h2 id="ffi_caveats">FFI caveats</h2>
632<p>
633The string buffer library has been designed to work well together with
634the FFI library. But due to the low-level nature of the FFI library,
635some care needs to be taken:
636</p>
637<p>
638First, please remember that FFI pointers are zero-indexed. The space
639returned by <tt>buf:reserve()</tt> and <tt>buf:ref()</tt> starts at the
640returned pointer and ends before <tt>len</tt> bytes after that.
641</p>
642<p>
643I.e. the first valid index is <tt>ptr[0]</tt> and the last valid index
644is <tt>ptr[len-1]</tt>. If the returned length is zero, there's no valid
645index at all. The returned pointer may even be <tt>NULL</tt>.
646</p>
647<p>
648The space pointed to by the returned pointer is only valid as long as
649the buffer is not modified in any way (neither append, nor consume, nor
650reset, etc.). The pointer is also not a GC anchor for the buffer object
651itself.
652</p>
653<p>
654Buffer data is only guaranteed to be byte-aligned. Casting the returned
655pointer to a data type with higher alignment may cause unaligned
656accesses. It depends on the CPU architecture whether this is allowed or
657not (it's always OK on x86/x64 and mostly OK on other modern
658architectures).
659</p>
660<p>
661FFI pointers or references do not count as GC anchors for an underlying
662object. E.g. an <tt>array</tt> allocated with <tt>ffi.new()</tt> is
663anchored by <tt>buf:set(array,&nbsp;len)</tt>, but not by
664<tt>buf:set(array+offset,&nbsp;len)</tt>. The addition of the offset
665creates a new pointer, even when the offset is zero. In this case, you
666need to make sure there's still a reference to the original array as
667long as its contents are in use by the buffer.
668</p>
669<p>
670Even though each LuaJIT VM instance is single-threaded (but you can
671create multiple VMs), FFI data structures can be accessed concurrently.
672Be careful when reading/writing FFI cdata from/to buffers to avoid
673concurrent accesses or modifications. In particular, the memory
674referenced by <tt>buf:set(cdata,&nbsp;len)</tt> must not be modified
675while buffer readers are working on it. Shared, but read-only memory
676mappings of files are OK, but only if the file does not change.
677</p>
678<br class="flush">
679</div>
680<div id="foot">
681<hr class="hide">
682Copyright &copy; 2005-2026
683<span class="noprint">
684&middot;
685<a href="contact.html">Contact</a>
686</span>
687</div>
688</body>
689</html>
generated by cgit v1.2.3-55-g6feb (git 2.39.1) at 2026-06-03 03:40:18 +0000