duel-postgresql 0.2.0 → 0.3.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
checksums.yaml CHANGED
@@ -1,7 +1,7 @@
1
1
  ---
2
2
  SHA256:
3
- metadata.gz: 621842b079d9c63ba899bde389ca544c7ca0773b85397db9d18dbd082fcd5f61
4
- data.tar.gz: 073a4fe8c7409169e5f06689329305fc433717a21af3688556b5172035ff6857
3
+ metadata.gz: bef0bbe4c1ac2f94456bac51aaec91704c843192e5455ad9e8df674deb81dea3
4
+ data.tar.gz: c66d907b9d22d687e425c5e4078465e7ec19c2155d1a457403d54fd3831e381a
5
5
  SHA512:
6
- metadata.gz: 4304d75eb5385232458f4f164d70ccde71276802047a3b02c1349494a1ba70521073b76430b4a653e52e46eea21f895feadcb52d8bac15402e22bd268a9717a6
7
- data.tar.gz: 61d2cae4af789f5264f569c48208231e6522dd19212e188535c97d0ef7d1dd4aa80e37c7b5f3b7a635508d2146492865efa18bc049ca325c99f49b8069bf3d53
6
+ metadata.gz: adeee8f3edcd810eda25673621c6019ecee27a17d6fb0ce4ecc82e2a85a249df315413743b59deca573076ffd657fe73422f526549986f498b799ac19ee7d502
7
+ data.tar.gz: 46503043f91a56cd3554b415ed3d479fa7b25cc9da2b8175da134ff493e68c1f571e81ea8515c7576c56379f25e0333d57a7fd6dffbde29224d12edbc7841780
data/CHANGELOG.md CHANGED
@@ -7,6 +7,36 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
7
7
 
8
8
  ## [Unreleased]
9
9
 
10
+ ## [0.3.0] - 2026-11-20
11
+
12
+ ### Changed
13
+
14
+ - `postgresql_blocks`' settings say `size_samples: false`, as Duel 0.3.0
15
+ names `sizes_samples:`.
16
+ - Its measures are named after the plugin, as every other Duel
17
+ plugin's are (`mysql_*`, `redis_*`): `postgresql_queries`,
18
+ `postgresql_rows`, `postgresql_blocks` and `postgresql_time`, in
19
+ place of `postgres_queries`, `postgres_rows`, `postgres_blocks` and
20
+ `postgres_time`. A `.duel.rb` that tunes one
21
+ (`c.measure("postgres_time", limit: 1.1)`) needs the new name.
22
+ - `sql:` is what each worker's database runs once its server starts
23
+ (a schema, say), in place of `setup:`, which is a hook's name
24
+ everywhere else in Duel.
25
+ - `server:` names the server to run, its `postgres` (a path, or a name
26
+ on the `PATH`), as the other server plugins' `server:` does, in place
27
+ of `bin:`, its directory, and `DUEL_POSTGRESQL_SERVER` in place of
28
+ `DUEL_POSTGRESQL_BIN`. Its other programs (`initdb`, `pg_ctl`) are
29
+ next to it.
30
+ - `tcp: true` starts each worker's server on a port of 127.0.0.1's
31
+ alone, with no Unix socket, as the other server plugins' does, and is
32
+ the default on Windows. It was a side effect of `pg_ctl: true`, which
33
+ only says what starts the server now.
34
+ - `server:` is found by Duel's `Duel.find_program`, as duel-mysql's
35
+ is: one with a `\` in it is a path off Windows too, and a name on the
36
+ `PATH` is taken as it is as well as with `.exe`.
37
+ - It requires Duel 0.3.0 or later (`~> 0.3.0`), the release it's
38
+ tested with, as Duel 0.3.0 changes what plugins and benchmarks call.
39
+
10
40
  ## [0.2.0] - 2026-11-16
11
41
 
12
42
  ### Added
@@ -128,7 +158,8 @@ later, with `pg_stat_statements`.
128
158
  CI.
129
159
  - RBS signatures in sig/, checked with Steep.
130
160
 
131
- [Unreleased]: https://github.com/sferik/duel-ruby/compare/duel-postgresql-v0.2.0...HEAD
161
+ [Unreleased]: https://github.com/sferik/duel-ruby/compare/duel-postgresql-v0.3.0...HEAD
162
+ [0.3.0]: https://github.com/sferik/duel-ruby/compare/duel-postgresql-v0.2.0...duel-postgresql-v0.3.0
132
163
  [0.2.0]: https://github.com/sferik/duel-ruby/compare/duel-postgresql-v0.1.8...duel-postgresql-v0.2.0
133
164
  [0.1.8]: https://github.com/sferik/duel-ruby/compare/duel-postgresql-v0.1.7...duel-postgresql-v0.1.8
134
165
  [0.1.7]: https://github.com/sferik/duel-ruby/compare/duel-postgresql-v0.1.6...duel-postgresql-v0.1.7
data/README.md CHANGED
@@ -10,10 +10,10 @@ every benchmark is judged on what PostgreSQL did for it too.
10
10
 
11
11
  | Metric | What it is | A regression |
12
12
  |---|---|---|
13
- | `postgres_queries` | the statements it ran (`BEGIN` and `COMMIT` too) | any increase |
14
- | `postgres_rows` | the rows they returned or changed | any increase |
15
- | `postgres_blocks` | the shared buffer blocks they touched, cached or not | judged like CPU time |
16
- | `postgres_time` | the time it spent planning and running them, in µs | judged like CPU time |
13
+ | `postgresql_queries` | the statements it ran (`BEGIN` and `COMMIT` too) | any increase |
14
+ | `postgresql_rows` | the rows they returned or changed | any increase |
15
+ | `postgresql_blocks` | the shared buffer blocks they touched, cached or not | judged like CPU time |
16
+ | `postgresql_time` | the time it spent planning and running them, in µs | judged like CPU time |
17
17
 
18
18
  So a change that adds a query per record (an N+1), or fetches rows it
19
19
  doesn't use, fails even when its CPU time on the Ruby side doesn't
@@ -25,19 +25,19 @@ on paired ratios, not counted.
25
25
  A regression says which statements did it: under each version's
26
26
  numbers, the three that did the most of that measure (Duel's
27
27
  `details`), with how many times each ran an iteration for
28
- `postgres_queries`, and their shares of it for the rest:
28
+ `postgresql_queries`, and their shares of it for the rest:
29
29
 
30
30
  ```
31
- postgres_queries: 2 queries → 6 queries an iteration
32
- postgres_queries (candidate): an iteration: 5× SELECT count(*) FROM comments WHERE post_id = $1; 1× SELECT * FROM posts
33
- postgres_queries (baseline): an iteration: 1× SELECT * FROM comments WHERE post_id = ANY($1); 1× SELECT * FROM posts
31
+ postgresql_queries: 2 queries → 6 queries an iteration
32
+ postgresql_queries (candidate): an iteration: 5× SELECT count(*) FROM comments WHERE post_id = $1; 1× SELECT * FROM posts
33
+ postgresql_queries (baseline): an iteration: 1× SELECT * FROM comments WHERE post_id = ANY($1); 1× SELECT * FROM posts
34
34
  ```
35
35
 
36
36
  They're of the sample the numbers came from, but for its `untimed`
37
37
  blocks, if it has any. Queries, rows and blocks are counted in their
38
38
  units (`1 query`, `12 rows`, though the summary line leaves out the one
39
- a name ends in: `postgres_rows 12→40`), and `postgres_time` is shown as time
40
- (`postgres_time: 412µs → 1.31ms an iteration`).
39
+ a name ends in: `postgresql_rows 12→40`), and `postgresql_time` is
40
+ shown as time (`postgresql_time: 412µs → 1.31ms an iteration`).
41
41
 
42
42
  ## Installation
43
43
 
@@ -63,10 +63,11 @@ Duel.configure { |c| c.use :postgresql }
63
63
  `c.use` requires the plugin, so `.duel.rb` doesn't need to.
64
64
 
65
65
  Each worker's server listens only on a Unix socket in the worker's own
66
- directory (or with `pg_ctl: true`, as by default on Windows, on a port
67
- of 127.0.0.1's alone, started and stopped by `pg_ctl`: there `postgres`
68
- won't run with an administrator's rights, which `pg_ctl` drops), with
69
- nothing written to disk that needn't be, and no
66
+ directory (or with `tcp: true`, as by default on Windows, on a port of
67
+ 127.0.0.1's alone). On Windows, `pg_ctl` starts and stops it, as
68
+ `postgres` won't run there with an administrator's rights, which
69
+ `pg_ctl` drops. It runs with nothing written to disk that needn't be,
70
+ and no
70
71
  autovacuum (which would change what a query reads, and its plan, while
71
72
  it's measured). Its URL is in `DUEL_POSTGRESQL_URL`, and in
72
73
  `DATABASE_URL` too, and libpq's `PGHOST`, `PGPORT`, `PGUSER` and
@@ -96,11 +97,12 @@ Options:
96
97
 
97
98
  ```ruby
98
99
  c.use :postgresql,
99
- bin: "/usr/lib/postgresql/17/bin", # its programs (default: DUEL_POSTGRESQL_BIN, or found)
100
- settings: {work_mem: "64MB"}, # more of its settings
101
- setup: File.read("db/structure.sql"), # SQL each worker's database runs first
102
- env: %w[DATABASE_URL PRIMARY_URL], # variables given its URL besides DUEL_POSTGRESQL_URL
103
- pg_ctl: true # started by pg_ctl, on a port of 127.0.0.1's (default: on Windows only)
100
+ server: "/usr/lib/postgresql/17/bin/postgres", # its postgres, initdb next to it (default: DUEL_POSTGRESQL_SERVER, or postgres)
101
+ settings: {work_mem: "64MB"}, # more of its settings, as work_mem=64MB (they win)
102
+ sql: File.read("db/structure.sql"), # SQL each worker's database runs first
103
+ env: %w[DATABASE_URL PRIMARY_URL], # variables given its URL besides DUEL_POSTGRESQL_URL (default: DATABASE_URL)
104
+ tcp: true, # on a port of 127.0.0.1's, not a Unix socket (default: on Windows only)
105
+ pg_ctl: true # started by pg_ctl (default: on Windows only)
104
106
  ```
105
107
 
106
108
  > [!WARNING]
@@ -109,16 +111,18 @@ c.use :postgresql,
109
111
  > fail, so give one of them other names (`env: %w[PRIMARY_URL]`, or
110
112
  > `env: []`).
111
113
 
112
- `bin` is found as the directory of the `postgres` on the `PATH`, links
113
- followed (Homebrew's `initdb` on the `PATH` may be libpq's, of another
114
- version), or Debian's newest under `/usr/lib/postgresql`.
114
+ Its other programs (`initdb`, `pg_ctl`) are the ones in the directory
115
+ of its `postgres` (`server:`, found as every Duel plugin finds its
116
+ server, by `Duel.find_program`), links followed (Homebrew's `initdb` on
117
+ the `PATH` may be libpq's, of another version). For a name that isn't
118
+ on the `PATH`, they're Debian's newest, under `/usr/lib/postgresql`.
115
119
 
116
120
  A database cluster takes `initdb`, and that's slow, so it's done once
117
121
  (Duel's `w.cached`): the first worker to start makes one, with the
118
122
  `duel` database and `pg_stat_statements` created in it, in your
119
123
  project's `tmp/duel/cache`, and every worker (in that run and the next
120
124
  ones) starts with a copy. One is made for each server's programs (and
121
- version of the plugin): another PostgreSQL gets its own. `duel --clean`
125
+ version of the plugin): another PostgreSQL gets its own. `duel clean`
122
126
  deletes them.
123
127
 
124
128
  The metrics come from `pg_stat_statements`, which PostgreSQL updates as
@@ -128,17 +132,17 @@ over a connection of the plugin's own, opened as each worker starts
128
132
  `c.metric`). `c.use` returns that connection, for a `.duel.rb` that
129
133
  reads the server too (`connection.value`, in a worker only). The
130
134
  reading's own statement is left out (by its query id), so there's
131
- nothing to take away (`calibrate: false`). `postgres_time` counts
135
+ nothing to take away (`calibrate: false`). `postgresql_time` counts
132
136
  planning too (`pg_stat_statements.track_planning`), as a change that
133
137
  stops reusing prepared statements plans every time. Up to 10,000
134
138
  distinct statements are kept: a benchmark that makes more (with `IN`
135
139
  lists of every length, say) loses the oldest, and its counts with them.
136
140
 
137
- `postgres_time` is the server's time on the clock, not CPU time, and
141
+ `postgresql_time` is the server's time on the clock, not CPU time, and
138
142
  another process's, so what else the machine does shows in it: it's
139
143
  judged for each benchmark, but left out of Duel's across-the-board
140
144
  check (`across: false`), whose margin is a quarter of its limit.
141
145
 
142
146
  Like any measure, each can be tuned or left out in `.duel.rb`:
143
- `c.measure("postgres_time", limit: 1.1)`,
144
- `c.measure("postgres_rows", skip: true)`.
147
+ `c.measure("postgresql_time", limit: 1.1)`,
148
+ `c.measure("postgresql_rows", skip: true)`.
@@ -2,6 +2,6 @@
2
2
 
3
3
  module Duel
4
4
  module PostgreSQL
5
- VERSION = "0.2.0"
5
+ VERSION = "0.3.0"
6
6
  end
7
7
  end
@@ -3,19 +3,18 @@
3
3
  require "duel"
4
4
  require "digest"
5
5
  require "fileutils"
6
- require "rbconfig"
7
6
  require_relative "postgresql/version"
8
7
 
9
8
  module Duel
10
9
  # A PostgreSQL server of each worker's own, and what it did for each
11
10
  # benchmark, from pg_stat_statements:
12
11
  #
13
- # - postgres_queries: the statements it ran (counted)
14
- # - postgres_rows: the rows they returned or changed (counted)
15
- # - postgres_blocks: the shared buffer blocks they touched, from the
12
+ # - postgresql_queries: the statements it ran (counted)
13
+ # - postgresql_rows: the rows they returned or changed (counted)
14
+ # - postgresql_blocks: the shared buffer blocks they touched, from the
16
15
  # cache or not (judged like CPU time: nearly steady, but a table
17
16
  # that grows, say, takes more)
18
- # - postgres_time: the time it spent planning and running them, in
17
+ # - postgresql_time: the time it spent planning and running them, in
19
18
  # microseconds (judged like CPU time)
20
19
  #
21
20
  # Where one regressed, it says which statements did the most of it, on
@@ -65,51 +64,51 @@ module Duel
65
64
  TCP = {"listen_addresses" => "127.0.0.1", "unix_socket_directories" => ""}.freeze
66
65
 
67
66
  # Configures Duel for the plugin (c.use :postgresql): the server and
68
- # its metrics. +bin+ is the directory of its programs (postgres and
69
- # initdb: by default, the real directory of the postgres on the
70
- # PATH), +settings+ more of its settings, +setup+ SQL each worker's
71
- # database runs when it's started (a schema, say), and +env+ the
72
- # variables given its URL besides DUEL_POSTGRESQL_URL. A cluster is
73
- # initdb'd once (Resource#cached), and each worker's is a copy.
67
+ # its metrics. +server+ is its postgres (a name on the PATH, or a
68
+ # path), whose directory has its other programs (initdb and pg_ctl),
69
+ # +settings+ more of its settings, +sql+ what each worker's database
70
+ # runs when it's started (a schema, say), and +env+ the variables
71
+ # given its URL besides DUEL_POSTGRESQL_URL. A cluster is initdb'd
72
+ # once (Resource#cached), and each worker's is a copy.
74
73
  #
75
- # With +pg_ctl+ (by default on Windows), the server is started (and
76
- # stopped) by pg_ctl, on a port of 127.0.0.1's: on Windows, postgres
77
- # won't run with an administrator's rights (CI's runners have them),
78
- # which pg_ctl drops, and its Unix sockets are new.
74
+ # Each worker's server listens on a Unix socket in its directory, or
75
+ # with +tcp+ (by default where there are no Unix sockets: Windows), on
76
+ # a port of 127.0.0.1's alone. With +pg_ctl+ (by default on Windows),
77
+ # it's started (and stopped) by pg_ctl: on Windows, postgres won't run
78
+ # with an administrator's rights (CI's runners have them), which
79
+ # pg_ctl drops.
79
80
  #
80
81
  # Returns the plugin's own connection to it, a worker's (see
81
82
  # Config#in_worker, and in tests, Duel::Testing.in_worker).
82
- def self.configure(config, bin: ENV.fetch("DUEL_POSTGRESQL_BIN", nil), settings: {}, setup: nil, env: %w[DATABASE_URL],
83
- pg_ctl: Duel.windows?)
83
+ def self.configure(config, server: ENV.fetch("DUEL_POSTGRESQL_SERVER", "postgres"), settings: {}, sql: nil, env: %w[DATABASE_URL],
84
+ tcp: !Duel.unix_sockets?, pg_ctl: Duel.windows?)
84
85
  config.prepare(:worker) do |w|
85
86
  require_pg
86
- programs = File.expand_path(bin || Cluster.find_bin ||
87
- w.failed("no postgres on the PATH (give its directory as bin:, or DUEL_POSTGRESQL_BIN)"))
87
+ programs = Cluster.programs(server) ||
88
+ w.fail!("no #{server}#{" on the PATH" unless Duel.path?(server)} (give postgres's path as server:, or DUEL_POSTGRESQL_SERVER)")
88
89
  data = w.path("data")
89
90
  cluster = Cluster.new(w, programs, pg_ctl: pg_ctl)
90
91
  cluster.create(data)
92
+ host, port, listen = tcp ? ["127.0.0.1", w.free_port, TCP] : [w.dir, PORT, {"unix_socket_directories" => w.dir}]
93
+ # (+settings+' last: of a setting given twice, postgres takes the
94
+ # last.)
95
+ all = SETTINGS.merge(listen, "port" => port).merge(settings)
91
96
  if pg_ctl
92
- host = "127.0.0.1"
93
- port = w.port
94
97
  # (Its settings in its postgresql.conf, as pg_ctl takes no
95
- # others whole. +settings+' last, which it takes.)
96
- cluster.start(data, SETTINGS.merge(TCP, "port" => port).merge(settings))
98
+ # others whole.)
99
+ cluster.start(data, all)
97
100
  w.at_stop { cluster.stop(data) }
98
101
  else
99
- host = w.dir
100
- port = PORT
101
102
  # (Stopped by INT: its clients are disconnected, not waited for, as
102
103
  # TERM would.)
103
- # (Of a setting given twice, postgres takes the last: +settings+'.)
104
- w.spawn(File.join(programs, "postgres"), "-D", data, "-k", w.dir, "-p", PORT.to_s,
105
- *SETTINGS.merge(settings).flat_map { |name, value| ["-c", "#{name}=#{value}"] }, stop_signal: "INT")
104
+ w.spawn(File.join(programs, "postgres"), "-D", data, *all.flat_map { |name, value| ["-c", "#{name}=#{value}"] }, stop_signal: "INT")
106
105
  end
107
106
  # (The user in the query, with no authority at all: Ruby's URI, which
108
107
  # Rails parses DATABASE_URL with, takes a user without a host for
109
108
  # a malformed host.)
110
109
  url = "postgres:///#{DATABASE}?host=#{host}&port=#{port}&user=#{USER}"
111
110
  w.wait_until { PG::Connection.ping(url) == PG::PQPING_OK }
112
- run(url, setup) if setup
111
+ run(url, sql) if sql
113
112
  {URL => url, "PGHOST" => host, "PGPORT" => port.to_s, "PGUSER" => USER, "PGDATABASE" => DATABASE}
114
113
  .merge(env.to_h { |name| [name, url] })
115
114
  end
@@ -151,10 +150,10 @@ module Duel
151
150
  # too small for it. Its readings leave themselves out (STATEMENTS), so
152
151
  # there's nothing to take away.
153
152
  METRICS = {
154
- "postgres_queries" => {kind: :counted, unit: %w[query queries], calibrate: false},
155
- "postgres_rows" => {kind: :counted, unit: %w[row rows], calibrate: false},
156
- "postgres_blocks" => {kind: :ratio, unit: %w[block blocks], floor: 100, sizes_samples: false},
157
- "postgres_time" => {kind: :ratio, unit: "µs", floor: 300, across: false}
153
+ "postgresql_queries" => {kind: :counted, unit: %w[query queries], calibrate: false},
154
+ "postgresql_rows" => {kind: :counted, unit: %w[row rows], calibrate: false},
155
+ "postgresql_blocks" => {kind: :timed, unit: %w[block blocks], floor: 100, size_samples: false},
156
+ "postgresql_time" => {kind: :timed, unit: "µs", floor: 300, across: false}
158
157
  }
159
158
 
160
159
  # What the worker's server has done so far, by statement (query id),
@@ -286,7 +285,7 @@ module Duel
286
285
  # A statement's numbers, or several's added up, by measure (time in
287
286
  # µs).
288
287
  def self.by_measure(numbers)
289
- NAMES.each_with_index.to_h { |name, i| [name, (name == "postgres_time") ? numbers.fetch(i).fdiv(1000) : numbers.fetch(i)] }
288
+ NAMES.each_with_index.to_h { |name, i| [name, (name == "postgresql_time") ? numbers.fetch(i).fdiv(1000) : numbers.fetch(i)] }
290
289
  end
291
290
 
292
291
  # Read over +connection+, the plugin's own (a worker's).
@@ -393,18 +392,20 @@ module Duel
393
392
  # A worker's database cluster: a copy of one initdb'd once, with the
394
393
  # database and pg_stat_statements created in it.
395
394
  class Cluster
396
- # A program's file ends with +EXE+: on Windows, .exe.
397
- EXE = RbConfig::CONFIG["EXEEXT"]
398
-
399
- # The real directory of the postgres on +path+ (where its initdb is:
400
- # the one on the PATH may be another version's, libpq's, say), or of
401
- # the newest Debian's; nil if there's none. (Its file is postgres
402
- # +exe+: postgres.exe, on Windows.)
403
- def self.find_bin(path = ENV.fetch("PATH", ""), debian: "/usr/lib/postgresql/*/bin", exe: EXE)
404
- postgres = path.split(File::PATH_SEPARATOR).map { |dir| File.join(dir, "postgres#{exe}") }.find { |file| File.executable?(file) }
395
+ # Debian's programs, by version (/usr/lib/postgresql/17/bin).
396
+ DEBIAN = "/usr/lib/postgresql/*/bin"
397
+
398
+ # The directory of +server+'s programs (see PostgreSQL.configure):
399
+ # the real directory, links followed, of the postgres
400
+ # Duel.find_program finds, a path's or a name's on +path+ (where its
401
+ # initdb is: the one on the PATH may be another version's, libpq's,
402
+ # say), or for a name it doesn't find, the newest Debian's
403
+ # (+debian+). Nil if there's none.
404
+ def self.programs(server, path = ENV.fetch("PATH", ""), debian: DEBIAN)
405
+ postgres = Duel.find_program(server, path)
405
406
  return File.dirname(File.realpath(postgres)) if postgres
406
407
 
407
- Dir[debian].max_by { |dir| File.basename(File.dirname(dir)).to_i }
408
+ Dir[debian].max_by { |dir| File.basename(File.dirname(dir)).to_i } unless Duel.path?(server)
408
409
  end
409
410
 
410
411
  # Where pg_ctl's cluster is in its build (see #make).
@@ -447,7 +448,7 @@ module Duel
447
448
  begin
448
449
  run("pg_ctl", "start", "-D", data, "-w", "-l", log)
449
450
  rescue Failed
450
- @resource.failed("postgres didn't start", log: log)
451
+ @resource.fail!("postgres didn't start", log: log)
451
452
  end
452
453
  end
453
454
 
@@ -460,7 +461,8 @@ module Duel
460
461
  # What a cluster is good for: these programs (until they change),
461
462
  # this version of the plugin, and how it's made (with pg_ctl or not).
462
463
  def key
463
- postgres = File.realpath(File.join(@bin, "postgres#{EXE}"))
464
+ # (postgres.exe, on Windows: there's one, as .programs found it.)
465
+ postgres = File.realpath(Duel.find_program("postgres", @bin)) # steep:ignore ArgumentTypeMismatch
464
466
  made = [postgres, File.mtime(postgres).to_f, *(PG_CTL if @pg_ctl)]
465
467
  "postgresql-#{VERSION}-#{Digest::SHA256.hexdigest(made.join("\0"))[0, 16]}"
466
468
  end
@@ -483,7 +485,7 @@ module Duel
483
485
  def make_with_pg_ctl(dir)
484
486
  data = File.join(dir, PG_CTL)
485
487
  initdb(data)
486
- port = @resource.port
488
+ port = @resource.free_port
487
489
  start(data, TCP.merge("port" => port))
488
490
  begin
489
491
  url = "postgres://#{USER}@127.0.0.1:#{port}"
@@ -1,7 +1,7 @@
1
1
  module Duel
2
2
  # A PostgreSQL server of each worker's own, and what it did for each
3
- # benchmark (postgres_queries, postgres_rows, postgres_blocks,
4
- # postgres_time).
3
+ # benchmark (postgresql_queries, postgresql_rows, postgresql_blocks,
4
+ # postgresql_time).
5
5
  module PostgreSQL
6
6
  VERSION: String
7
7
 
@@ -39,14 +39,14 @@ module Duel
39
39
  type numbers = Array[Integer]
40
40
 
41
41
  # Configures Duel for the plugin (c.use :postgresql): the server
42
- # (+bin+ the directory of its programs, +settings+ more of its
43
- # settings, +setup+ SQL each worker's database runs when it's started,
44
- # +env+ the variables besides DUEL_POSTGRESQL_URL given its URL) and
45
- # its metrics. Returns the plugin's own connection to it, a worker's.
42
+ # (+server+ its postgres, its other programs next to it, +settings+
43
+ # more of its settings, +sql+ what each worker's database runs when
44
+ # it's started, +env+ the variables besides DUEL_POSTGRESQL_URL given
45
+ # its URL) and its metrics. Returns the plugin's own connection to it, a worker's.
46
46
  # +pg_ctl+: started and stopped by pg_ctl, on a port of 127.0.0.1's
47
47
  # (by default on Windows).
48
- def self.configure: (Config config, ?bin: String?, ?settings: Hash[String | Symbol, String | Integer], ?setup: String?,
49
- ?env: Array[String], ?pg_ctl: bool) -> InWorker[::PG::Connection]
48
+ def self.configure: (Config config, ?server: String, ?settings: Hash[String | Symbol, String | Integer], ?sql: String?,
49
+ ?env: Array[String], ?tcp: bool, ?pg_ctl: bool) -> InWorker[::PG::Connection]
50
50
 
51
51
  # Runs SQL on a connection of its own to +url+; returns the last
52
52
  # statement's result.
@@ -157,12 +157,13 @@ module Duel
157
157
  # Where pg_ctl's cluster is in its build.
158
158
  PG_CTL: String
159
159
 
160
- # The real directory of the postgres on +path+, or of the newest
161
- # Debian's; nil if there's none.
162
- # A program's file ends with +EXE+ (.exe, on Windows).
163
- EXE: String
160
+ # Debian's programs, by version.
161
+ DEBIAN: String
164
162
 
165
- def self.find_bin: (?String path, ?debian: String, ?exe: String) -> String?
163
+ # The directory of a server's programs: the real directory of the
164
+ # postgres at a path, or by a name on +path+, or the newest
165
+ # Debian's. Nil if there's none.
166
+ def self.programs: (String server, ?String path, ?debian: String) -> String?
166
167
 
167
168
  def initialize: (Resource resource, String bin, ?pg_ctl: bool) -> void
168
169
 
metadata CHANGED
@@ -1,7 +1,7 @@
1
1
  --- !ruby/object:Gem::Specification
2
2
  name: duel-postgresql
3
3
  version: !ruby/object:Gem::Version
4
- version: 0.2.0
4
+ version: 0.3.0
5
5
  platform: ruby
6
6
  authors:
7
7
  - Erik Berlin
@@ -16,7 +16,7 @@ dependencies:
16
16
  -
17
17
  - ~>
18
18
  - !ruby/object:Gem::Version
19
- version: "0.2"
19
+ version: 0.3.0
20
20
  type: :runtime
21
21
  prerelease: false
22
22
  version_requirements: !ruby/object:Gem::Requirement
@@ -24,7 +24,7 @@ dependencies:
24
24
  -
25
25
  - ~>
26
26
  - !ruby/object:Gem::Version
27
- version: "0.2"
27
+ version: 0.3.0
28
28
  - !ruby/object:Gem::Dependency
29
29
  name: pg
30
30
  requirement: !ruby/object:Gem::Requirement