Tarantool development patches archive
 help / color / mirror / Atom feed
From: Vladimir Davydov <vdavydov.dev@gmail.com>
To: kostja@tarantool.org
Cc: tarantool-patches@freelists.org
Subject: Re: [PATCH 07/12] vinyl: sanitize full/empty key stmt detection
Date: Fri, 1 Mar 2019 15:57:36 +0300	[thread overview]
Message-ID: <20190301125736.bppcf26bt3xvxozx@esperanza> (raw)
In-Reply-To: <b89cd218b341a05d5f4d28f7aeb43df594433d4e.1550744027.git.vdavydov.dev@gmail.com>

On Thu, Feb 21, 2019 at 01:26:07PM +0300, Vladimir Davydov wrote:
> Historically, we use tuple_field_count to check whether a statement
> represents an empty key (match all) or a full key (point lookup): if
> the number of fields in a tuple is greater than or equal to the number
> of parts in a key definition, it can be used as a full key; if the
> number of fields is zero, then the statement represents an empty key.
> 
> While this used to be correct not so long ago, appearance of JSON
> indexes changed the rules of the game: now a tuple can have nested
> indexed fields so that the same field number appears in the key
> definition multiple times. This means tuple_field_count can be less
> than the number of key parts and hence the full key check won't work
> for a statement representing a tuple.
> 
> Actually, any tuple in vinyl can be used as a full key as it has all
> key parts by definition, there's no need to use tuple_field_count for
> such statements - we only need to do that for statements representing
> keys. Keeping that in mind, let's introduce helpers for checking
> whether a statement can be used as a full/empty key and use them
> throughout the code.
> ---
>  src/box/vinyl.c            |  2 +-
>  src/box/vy_cache.c         | 14 +++++++++-----
>  src/box/vy_mem.c           |  2 +-
>  src/box/vy_point_lookup.c  |  5 ++---
>  src/box/vy_range.c         |  5 ++---
>  src/box/vy_read_iterator.c |  6 +++---
>  src/box/vy_read_set.c      | 18 ++++++------------
>  src/box/vy_run.c           |  2 +-
>  src/box/vy_stmt.h          | 38 ++++++++++++++++++++++++++++++++++++++
>  src/box/vy_tx.c            |  4 ++--
>  10 files changed, 65 insertions(+), 31 deletions(-)

Pushed to 2.1 and 1.10.

Also, implemented a test that demonstrates the issues with JSON indexes
in vinyl and pushed it to 2.1:

From 2f14800131340621c0818fc326ea618aa4296c63 Mon Sep 17 00:00:00 2001
From: Vladimir Davydov <vdavydov.dev@gmail.com>
Date: Fri, 1 Mar 2019 15:45:32 +0300
Subject: [PATCH] test: check vinyl/json corner cases

Follow-up

  5993e149d90e vinyl: sanitize full/empty key stmt detection
  4273ec52e122 box: introduce JSON Indexes

diff --git a/test/vinyl/json.result b/test/vinyl/json.result
new file mode 100644
index 00000000..f17619f4
--- /dev/null
+++ b/test/vinyl/json.result
@@ -0,0 +1,141 @@
+test_run = require('test_run').new()
+---
+...
+--
+-- Lookup in the primary index when applying a deferred DELETE
+-- for a secondary index on commit.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+---
+...
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+---
+...
+sk = s:create_index('sk', {unique = false, parts = {2, 'unsigned'}})
+---
+...
+s:replace{{a = 1, b = 2, c = 3}, 10}
+---
+- [{'b': 2, 'a': 1, 'c': 3}, 10]
+...
+sk:select()
+---
+- - [{'b': 2, 'a': 1, 'c': 3}, 10]
+...
+s:drop()
+---
+...
+--
+-- Lookup on INSERT to check the unique constraint.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+---
+...
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+---
+...
+s:replace{{a = 1, b = 2, c = 3}, 1}
+---
+- [{'b': 2, 'a': 1, 'c': 3}, 1]
+...
+box.snapshot()
+---
+- ok
+...
+s:replace{{a = 1, b = 2, c = 3}, 2}
+---
+- [{'b': 2, 'a': 1, 'c': 3}, 2]
+...
+s:insert{{a = 1, b = 2, c = 3}, 3}
+---
+- error: Duplicate key exists in unique index 'pk' in space 'test'
+...
+pk:stat().disk.iterator.lookup -- 0 (served from memory)
+---
+- 0
+...
+s:drop()
+---
+...
+--
+-- Gap locks coalescing.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+---
+...
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+---
+...
+s:replace{{a = 1, b = 1, c = 1}}
+---
+- [{'b': 1, 'a': 1, 'c': 1}]
+...
+s:replace{{a = 1, b = 1, c = 2}}
+---
+- [{'b': 1, 'a': 1, 'c': 2}]
+...
+box.begin()
+---
+...
+gap_locks_1 = box.stat.vinyl().tx.gap_locks
+---
+...
+s:select({1, 1}, {iterator = 'ge', limit = 1})
+---
+- - [{'b': 1, 'a': 1, 'c': 1}]
+...
+s:select({1, 1}, {iterator = 'gt'})
+---
+- []
+...
+gap_locks_2 = box.stat.vinyl().tx.gap_locks
+---
+...
+gap_locks_2 - gap_locks_1 -- 2 (tracking intervals must not be coalesced)
+---
+- 2
+...
+box.commit()
+---
+...
+s:drop()
+---
+...
+--
+-- Cache iterator stop condition.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+---
+...
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+---
+...
+s:replace{{a = 1, b = 1, c = 1}}
+---
+- [{'b': 1, 'a': 1, 'c': 1}]
+...
+s:replace{{a = 1, b = 1, c = 2}}
+---
+- [{'b': 1, 'a': 1, 'c': 2}]
+...
+s:replace{{a = 1, b = 1, c = 3}}
+---
+- [{'b': 1, 'a': 1, 'c': 3}]
+...
+s:insert{{a = 1, b = 1, c = 3}}
+---
+- error: Duplicate key exists in unique index 'pk' in space 'test'
+...
+s:select{1, 1, 1}
+---
+- - [{'b': 1, 'a': 1, 'c': 1}]
+...
+s:select{1, 1}
+---
+- - [{'b': 1, 'a': 1, 'c': 1}]
+  - [{'b': 1, 'a': 1, 'c': 2}]
+  - [{'b': 1, 'a': 1, 'c': 3}]
+...
+s:drop()
+---
+...
diff --git a/test/vinyl/json.test.lua b/test/vinyl/json.test.lua
new file mode 100644
index 00000000..2f9f9f6e
--- /dev/null
+++ b/test/vinyl/json.test.lua
@@ -0,0 +1,53 @@
+test_run = require('test_run').new()
+
+--
+-- Lookup in the primary index when applying a deferred DELETE
+-- for a secondary index on commit.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+sk = s:create_index('sk', {unique = false, parts = {2, 'unsigned'}})
+s:replace{{a = 1, b = 2, c = 3}, 10}
+sk:select()
+s:drop()
+
+--
+-- Lookup on INSERT to check the unique constraint.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+s:replace{{a = 1, b = 2, c = 3}, 1}
+box.snapshot()
+s:replace{{a = 1, b = 2, c = 3}, 2}
+s:insert{{a = 1, b = 2, c = 3}, 3}
+pk:stat().disk.iterator.lookup -- 0 (served from memory)
+s:drop()
+
+--
+-- Gap locks coalescing.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+s:replace{{a = 1, b = 1, c = 1}}
+s:replace{{a = 1, b = 1, c = 2}}
+box.begin()
+gap_locks_1 = box.stat.vinyl().tx.gap_locks
+s:select({1, 1}, {iterator = 'ge', limit = 1})
+s:select({1, 1}, {iterator = 'gt'})
+gap_locks_2 = box.stat.vinyl().tx.gap_locks
+gap_locks_2 - gap_locks_1 -- 2 (tracking intervals must not be coalesced)
+box.commit()
+s:drop()
+
+--
+-- Cache iterator stop condition.
+--
+s = box.schema.space.create('test', {engine = 'vinyl'})
+pk = s:create_index('pk', {parts = { {'[1].a', 'unsigned'}, {'[1].b', 'unsigned'}, {'[1].c', 'unsigned'} }})
+s:replace{{a = 1, b = 1, c = 1}}
+s:replace{{a = 1, b = 1, c = 2}}
+s:replace{{a = 1, b = 1, c = 3}}
+s:insert{{a = 1, b = 1, c = 3}}
+s:select{1, 1, 1}
+s:select{1, 1}
+s:drop()

  parent reply	other threads:[~2019-03-01 12:57 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-02-21 10:26 [PATCH 00/12] vinyl: do not fill secondary tuples with nulls Vladimir Davydov
2019-02-21 10:26 ` [PATCH 01/12] vinyl: use vy_lsm_env::empty_key where appropriate Vladimir Davydov
2019-02-21 10:59   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 02/12] vinyl: make vy_tuple_delete static Vladimir Davydov
2019-02-21 11:00   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 03/12] key_def: cleanup virtual function initialization Vladimir Davydov
2019-02-21 11:01   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 12:05     ` Vladimir Davydov
2019-02-21 10:26 ` [PATCH 04/12] key_def: move cmp and hash functions declarations to key_def.h Vladimir Davydov
2019-02-21 11:02   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 05/12] vinyl: move vy_tuple_key_contains_null to generic code Vladimir Davydov
2019-02-21 11:02   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 06/12] vinyl: move vy_key_dup " Vladimir Davydov
2019-02-21 11:04   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 11:52     ` Vladimir Davydov
2019-02-21 10:26 ` [PATCH 07/12] vinyl: sanitize full/empty key stmt detection Vladimir Davydov
2019-02-21 11:10   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 12:11     ` Vladimir Davydov
2019-03-01 12:57   ` Vladimir Davydov [this message]
2019-02-21 10:26 ` [PATCH 08/12] vinyl: remove optimized comparators Vladimir Davydov
2019-02-21 11:11   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 09/12] vinyl: introduce statement environment Vladimir Davydov
2019-02-21 11:14   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 10/12] vinyl: rename key stmt construction routine Vladimir Davydov
2019-02-21 11:15   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 12:14     ` Vladimir Davydov
2019-02-21 10:26 ` [PATCH 11/12] vinyl: don't use IPROTO_SELECT type for key statements Vladimir Davydov
2019-02-21 11:16   ` [tarantool-patches] " Konstantin Osipov
2019-02-21 10:26 ` [PATCH 12/12] vinyl: do not fill secondary tuples with nulls when decoded Vladimir Davydov
2019-02-21 15:39 ` [PATCH 00/12] vinyl: do not fill secondary tuples with nulls Vladimir Davydov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190301125736.bppcf26bt3xvxozx@esperanza \
    --to=vdavydov.dev@gmail.com \
    --cc=kostja@tarantool.org \
    --cc=tarantool-patches@freelists.org \
    --subject='Re: [PATCH 07/12] vinyl: sanitize full/empty key stmt detection' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox