Commit0a8de93

committed

Speed up lexing of long JSON strings

Use optimized linear search when looking ahead for end quotes,backslashes, and non-printable characters. This results in nearly 40%faster JSON parsing on x86-64 when most values are long strings, andall platforms should see some improvement.Reviewed by Andres Freund and Nathan BossartDiscussion:https://www.postgresql.org/message-id/CAFBsxsGhaR2KQ5eisaK%3D6Vm60t%3DaxhD8Ckj1qFoCH1pktZi%2B2w%40mail.gmail.comDiscussion:https://www.postgresql.org/message-id/CAFBsxsESLUyJ5spfOSyPrOvKUEYYNqsBosue9SV1j8ecgNXSKA%40mail.gmail.com

1 parent0551912 commit0a8de93Copy full SHA for 0a8de93

File tree

+28

-3

lines changed

+28

-3

lines changed

Lines changed: 10 additions & 3 deletions

Original file line number	Diff line number	Diff line change
`@@ -19,6 +19,7 @@`
`19`	`19`
`20`	`20`	`#include"common/jsonapi.h"`
`21`	`21`	`#include"mb/pg_wchar.h"`
	`22`	`+#include"port/pg_lfind.h"`
`22`	`23`
`23`	`24`	`#ifndefFRONTEND`
`24`	`25`	`#include"miscadmin.h"`
`@@ -844,7 +845,7 @@ json_lex_string(JsonLexContext *lex)`
`844`	`845`	`}`
`845`	`846`	`else`
`846`	`847`	`{`
`847`		`-char*p;`
	`848`	`+char*p=s;`
`848`	`849`
`849`	`850`	`if (hi_surrogate!=-1)`
`850`	`851`	`returnJSON_UNICODE_LOW_SURROGATE;`
`@@ -853,11 +854,17 @@ json_lex_string(JsonLexContext *lex)`
`853`	`854`	`* Skip to the first byte that requires special handling, so we`
`854`	`855`	`* can batch calls to appendBinaryStringInfo.`
`855`	`856`	`*/`
`856`		`-for (p=s;p<end;p++)`
	`857`	`+while (p<end-sizeof(Vector8)&&`
	`858`	`+ !pg_lfind8('\\', (uint8*)p,sizeof(Vector8))&&`
	`859`	`+ !pg_lfind8('"', (uint8*)p,sizeof(Vector8))&&`
	`860`	`+ !pg_lfind8_le(31, (uint8*)p,sizeof(Vector8)))`
	`861`	`+p+=sizeof(Vector8);`
	`862`	`+`
	`863`	`+for (;p<end;p++)`
`857`	`864`	`{`
`858`	`865`	`if (p=='\\'\|\|p=='"')`
`859`	`866`	`break;`
`860`		`-elseif ((unsignedchar)*p<32)`
	`867`	`+elseif ((unsignedchar)*p <=31)`
`861`	`868`	`{`
`862`	`869`	`/* Per RFC4627, these characters MUST be escaped. */`
`863`	`870`	`/*`

Lines changed: 13 additions & 0 deletions

Original file line number	Diff line number	Diff line change
`@@ -42,6 +42,19 @@ LINE 1: SELECT '"\v"'::json;`
`42`	`42`	`^`
`43`	`43`	`DETAIL: Escape sequence "\v" is invalid.`
`44`	`44`	`CONTEXT: JSON data, line 1: "\v...`
	`45`	`+-- Check fast path for longer strings (at least 16 bytes long)`
	`46`	`+SELECT ('"'\|\|repeat('.', 12)\|\|'abc"')::json; -- OK`
	`47`	`+ json`
	`48`	`+-------------------`
	`49`	`+ "............abc"`
	`50`	`+(1 row)`
	`51`	`+`
	`52`	`+SELECT ('"'\|\|repeat('.', 12)\|\|'abc\n"')::json; -- OK, legal escapes`
	`53`	`+ json`
	`54`	`+---------------------`
	`55`	`+ "............abc\n"`
	`56`	`+(1 row)`
	`57`	`+`
`45`	`58`	`-- see json_encoding test for input with unicode escapes`
`46`	`59`	`-- Numbers.`
`47`	`60`	`SELECT '1'::json;-- OK`

Lines changed: 5 additions & 0 deletions

Original file line number	Diff line number	Diff line change
`@@ -7,6 +7,11 @@ SELECT '"abc`
`7`	`7`	`def"'::json;-- ERROR, unescaped newline in string constant`
`8`	`8`	`SELECT'"\n\"\\"'::json;-- OK, legal escapes`
`9`	`9`	`SELECT'"\v"'::json;-- ERROR, not a valid JSON escape`
	`10`	`+`
	`11`	`+-- Check fast path for longer strings (at least 16 bytes long)`
	`12`	`+SELECT ('"'\|\|repeat('.',12)\|\|'abc"')::json;-- OK`
	`13`	`+SELECT ('"'\|\|repeat('.',12)\|\|'abc\n"')::json;-- OK, legal escapes`
	`14`	`+`
`10`	`15`	`-- see json_encoding test for input with unicode escapes`
`11`	`16`
`12`	`17`	`-- Numbers.`

Comments

(0)