1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
//! The array commands, on the wire.
//!
//! The same shape as [`super::lists`]: the name has been looked up and the
//! arity has been checked, so this turns arguments into a call on [`Keyspace`]
//! and the answer into a reply. No decisions about arrays are made here.
//!
//! # Every index is read before any of them is used
//!
//! `ARMGET`, `ARMSET`, `ARDEL` and `ARDELRANGE` all take a list of indices, and
//! all four of them read the whole list before they touch the key. That is
//! Redis's behaviour and it matters: `ARDEL k 1 2 nope` deletes nothing rather
//! than deleting two things and then failing, and `ARMGET k 1 nope` writes no
//! part of an array header. Doing it any other way leaves a client that got an
//! error unable to say what happened.
//!
//! # The two numbers
//!
//! `ARLEN` is the highest populated index plus one and `ARCOUNT` is how many
//! indices hold something. They are both unsigned, and `ARLEN` genuinely can be
//! a number that does not fit an `i64`, which is why the replies here go
//! through [`Out::uint`].
use yo_common::num::{DOUBLE_MAX, parse_i64, write_g17};
use yo_common::{Code, Error, Result};
use yo_kv::arrays::{Aggregate, Grep, Op, Test, parse_grep_bound, parse_index, parse_seek_index};
use yo_kv::{ArrayElement, Db, Keyspace};
use super::args::{self, Args};
use super::table::Spec;
use crate::reply::Out;
/// Run one array command.
///
/// Every command in the group names one key and names it first, so the stripe
/// is found once here and everything below goes on taking a keyspace.
pub(super) fn execute(db: &Db, spec: &Spec, args: Args<'_>, out: &mut Out) -> Result<()> {
let mut held = db.hold(args.get(1));
let db = &mut *held;
match spec.name {
"arset" => {
let index = parse_index(args.get(2))?;
let values = (3..args.len()).map(|i| args.get(i));
out.uint(db.arset(args.get(1), index, values)?);
}
"armset" => {
// Pairs, so an odd tail is an arity error and not a syntax one.
if !args.len().is_multiple_of(2) {
return Err(args::wrong_arity(spec.name));
}
for i in (2..args.len()).step_by(2) {
parse_index(args.get(i))?;
}
let pairs = (2..args.len())
.step_by(2)
.map(|i| (parse_index(args.get(i)).unwrap_or(0), args.get(i + 1)));
out.uint(db.armset(args.get(1), pairs)?);
}
"arget" => {
let index = index_after_type(db, args.get(1), args.get(2))?;
match db.arget(args.get(1), index)? {
Some(e) => element(out, e),
None => out.nil(),
}
}
"armget" => {
for i in 2..args.len() {
index_after_type(db, args.get(1), args.get(i))?;
}
out.array(args.len() - 2);
let indices = (2..args.len()).map(|i| parse_index(args.get(i)).unwrap_or(0));
db.arget_into(args.get(1), indices, |el| reply(out, el))?;
}
"argetrange" => {
let start = parse_index(args.get(2))?;
let end = parse_index(args.get(3))?;
// The header carries the count, and the count is known from the two
// ends before a single position is read, so this needs none of the
// mark and close dance the collection walks use.
let mark = out.len();
let mut n = 0;
let len = db.argetrange(args.get(1), start, end, |el| {
reply(out, el);
n += 1;
})?;
debug_assert_eq!(len, n);
out.close_array(mark, usize::try_from(n).unwrap_or(usize::MAX));
}
"arlen" => out.uint(db.arlen(args.get(1))?),
"arcount" => out.uint(db.arcount(args.get(1))?),
"ardel" => {
for i in 2..args.len() {
parse_index(args.get(i))?;
}
let indices = (2..args.len()).map(|i| parse_index(args.get(i)).unwrap_or(0));
out.uint(db.ardel(args.get(1), indices)?);
}
"ardelrange" => {
if !args.len().is_multiple_of(2) {
return Err(args::wrong_arity(spec.name));
}
for i in 2..args.len() {
parse_index(args.get(i))?;
}
let ranges = (2..args.len()).step_by(2).map(|i| {
(
parse_index(args.get(i)).unwrap_or(0),
parse_index(args.get(i + 1)).unwrap_or(0),
)
});
out.uint(db.ardelrange(args.get(1), ranges)?);
}
"arinsert" => {
let values = (2..args.len()).map(|i| args.get(i));
out.uint(db.arinsert(args.get(1), values)?);
}
"arring" => {
// Redis reads the size before it looks at the key, so a bad size
// against a string is a bad size and not a wrong type.
let size =
parse_i64(args.get(2)).ok_or_else(|| Error::new(Code::Invalid, "invalid size"))?;
if size <= 0 {
return Err(Error::new(Code::Invalid, "size must be positive"));
}
let values = (3..args.len()).map(|i| args.get(i));
out.uint(db.arring(args.get(1), size as u64, values)?);
}
"arnext" => match db.arnext(args.get(1))? {
Some(index) => out.uint(index),
// The cursor is at the top of the space and there is no next index
// to name, which is the one thing this command cannot answer with a
// number.
None => out.nil(),
},
"arseek" => {
let index = parse_seek_index(args.get(2))?;
out.uint(u64::from(db.arseek(args.get(1), index)?));
}
"arlastitems" => {
let count = args
.int(2)
.map_err(|_| Error::new(Code::Invalid, "invalid COUNT"))?;
// Nothing asked for is an empty reply, and Redis answers it before
// it has read the option or looked at the key, so this does too.
if count <= 0 {
out.array(0);
return Ok(());
}
let newest_first = match args.len() {
3 => false,
4 if args::is(args.get(3), b"REV") => true,
4 => return Err(args::syntax()),
_ => return Err(args::wrong_arity(spec.name)),
};
let mark = out.len();
let n = db.arlastitems(args.get(1), count as u64, newest_first, |el| reply(out, el))?;
out.close_array(mark, usize::try_from(n).unwrap_or(usize::MAX));
}
"arscan" => {
let start = parse_index(args.get(2))?;
let end = parse_index(args.get(3))?;
let limit = match args.len() {
4 => u64::MAX,
6 if args::is(args.get(4), b"LIMIT") => {
let n = args.int(5)?;
if n <= 0 {
return Err(Error::new(Code::Invalid, "LIMIT must be positive"));
}
n as u64
}
6 => return Err(args::syntax()),
_ => return Err(args::wrong_arity(spec.name)),
};
let mark = out.len();
let n = db.arscan(args.get(1), start, end, limit, |index, el| {
// A pair per element rather than a flat list, so a client can
// read the reply without knowing whether it asked for a limit.
out.array(2);
out.uint(index);
element(out, el);
})?;
out.close_array(mark, usize::try_from(n).unwrap_or(usize::MAX));
}
"argrep" => {
// Both bounds are read before the plan is, so `ARGREP k x 0 EXACT`
// is a bad index and not a syntax error, whichever way round the
// two mistakes are written.
let start = parse_grep_bound(args.get(2))?;
let end = parse_grep_bound(args.get(3))?;
let mut grep = Grep::new();
let (mut all, mut nocase, mut withvalues) = (false, false, false);
let mut limit = u64::MAX;
// One pass, so predicates and options mix freely and the last of a
// repeated option wins. That is Redis's parser and it is the reason
// `ARGREP k 0 1 NOCASE RE a` and `ARGREP k 0 1 RE a NOCASE` are the
// same command.
let mut i = 4;
while i < args.len() {
let token = args.get(i);
let test = match token {
t if args::is(t, b"EXACT") => Some(Test::Exact),
t if args::is(t, b"MATCH") => Some(Test::Match),
t if args::is(t, b"GLOB") => Some(Test::Glob),
t if args::is(t, b"RE") => Some(Test::Re),
_ => None,
};
if let Some(test) = test {
if i + 1 >= args.len() {
return Err(args::syntax());
}
grep.push(test, args.get(i + 1))?;
i += 2;
continue;
}
match token {
t if args::is(t, b"LIMIT") => {
if i + 1 >= args.len() {
return Err(args::syntax());
}
let n = args.int(i + 1)?;
if n <= 0 {
return Err(Error::new(Code::Invalid, "LIMIT must be positive"));
}
limit = n as u64;
i += 2;
}
t if args::is(t, b"AND") => {
all = true;
i += 1;
}
t if args::is(t, b"OR") => {
all = false;
i += 1;
}
t if args::is(t, b"WITHVALUES") => {
withvalues = true;
i += 1;
}
t if args::is(t, b"NOCASE") => {
nocase = true;
i += 1;
}
_ => return Err(args::syntax()),
}
}
// Asking for nothing is a syntax error rather than an empty reply,
// because a client that meant to send a predicate and lost it in a
// shell should hear about it.
if grep.is_empty() {
return Err(args::syntax());
}
grep.compile(all, nocase)?;
let mark = out.len();
let n = db.argrep(args.get(1), start, end, limit, &mut grep, |index, el| {
// A bare index unless the values were asked for, and a pair of
// index and value when they were. Either way one entry per hit,
// which is what the header at the end counts.
if withvalues {
out.array(2);
}
out.uint(index);
if withvalues {
element(out, el);
}
})?;
out.close_array(mark, usize::try_from(n).unwrap_or(usize::MAX));
}
"arop" => {
let start = parse_index(args.get(2))?;
let end = parse_index(args.get(3))?;
let op = match args.get(4) {
w if args::is(w, b"SUM") => Op::Sum,
w if args::is(w, b"MIN") => Op::Min,
w if args::is(w, b"MAX") => Op::Max,
w if args::is(w, b"AND") => Op::And,
w if args::is(w, b"OR") => Op::Or,
w if args::is(w, b"XOR") => Op::Xor,
w if args::is(w, b"MATCH") => Op::Match,
w if args::is(w, b"USED") => Op::Used,
_ => return Err(Error::new(Code::Invalid, "unknown operation")),
};
// MATCH is the only one that takes a value, and it says so in its
// own words rather than as an arity error.
if op == Op::Match {
if args.len() != 6 {
return Err(Error::new(Code::Invalid, "MATCH requires a value argument"));
}
} else if args.len() != 5 {
return Err(args::wrong_arity(spec.name));
}
match db.arop(args.get(1), start, end, op, args.get(5))? {
Aggregate::Int(n) => out.int(n),
Aggregate::Num(d) => {
// Redis prints an aggregate with seventeen significant
// digits and not with the shortest round trip printer every
// other reply uses, so this is the one place those two
// disagree.
let mut buf = [0u8; DOUBLE_MAX];
out.bulk(write_g17(&mut buf, d));
}
Aggregate::None => out.nil(),
}
}
"arinfo" => {
let full = match args.len() {
2 => false,
3 if args::is(args.get(2), b"FULL") => true,
_ => return Err(args::syntax()),
};
let info = db.arinfo(args.get(1), full)?;
out.map(if full { 12 } else { 7 });
out.bulk(b"count");
out.uint(info.count);
out.bulk(b"len");
out.uint(info.len);
out.bulk(b"next-insert-index");
out.uint(info.next_insert);
out.bulk(b"slices");
out.uint(info.slices);
out.bulk(b"directory-size");
out.uint(info.directory_size);
out.bulk(b"super-dir-entries");
// Redis grows a second level above the directory once one level is
// wasteful. We do not have one, and the number is here so that a
// client reading the map finds the field it expects (D-20).
out.uint(0);
out.bulk(b"slice-size");
out.uint(info.slice_size);
if full {
out.bulk(b"dense-slices");
out.uint(info.dense_slices);
out.bulk(b"sparse-slices");
out.uint(info.sparse_slices);
out.bulk(b"avg-dense-size");
out.double(info.avg_dense_size);
out.bulk(b"avg-dense-fill");
out.double(info.avg_dense_fill);
out.bulk(b"avg-sparse-size");
out.double(info.avg_sparse_size);
}
}
other => unreachable!("the table sent {other} to the array group"),
}
Ok(())
}
/// An index, but with the type of the key reported first when it is a bad one.
///
/// `ARGET` and `ARMGET` are the only two array commands that look the key up
/// before they read the index, so `ARGET stringkey -1` is a wrong type where
/// `ARSET stringkey -1 v` is a bad index. The difference is visible to a client
/// and there is no reasoning behind it beyond the order the two commands happen
/// to be written in, so this reproduces it without paying for it: the type is
/// only looked at once the index has already failed, and the ordinary path is
/// still one lookup.
fn index_after_type(db: &mut Keyspace, key: &[u8], bytes: &[u8]) -> Result<u64> {
match parse_index(bytes) {
Ok(index) => Ok(index),
Err(e) => {
db.arlen(key)?;
Err(e)
}
}
}
/// One element, or a null for a hole.
fn reply(out: &mut Out, el: Option<ArrayElement<'_>>) {
match el {
Some(e) => element(out, e),
None => out.nil(),
}
}
/// One element as the bulk string a client sees.
///
/// A value stored as a number is formatted here, into a stack buffer, and
/// copied once into the reply. Formatting it when it was stored would have cost
/// the same work on the write path and the bytes to hold it afterwards.
fn element(out: &mut Out, e: ArrayElement<'_>) {
match e {
ArrayElement::Str(s) => out.bulk(s),
ArrayElement::Short(ref s) => out.bulk(s.as_bytes()),
ArrayElement::Int(n) => out.bulk_int(n),
ArrayElement::Float(_) => {
let mut buf = [0u8; yo_kv::array::ELEMENT_MAX];
out.bulk(e.text(&mut buf));
}
}
}