Test LIKE with implicit \ escape
#13288
Conversation
| # if "%%" in the pattern was simplified to "%", the pattern semantics would change | ||
| # same as above query, but with data coming from a table, so that constant folding cannot kick in, but expression simplification can | ||
| query TB | ||
| SELECT a, a LIKE '\%%' FROM inputs | ||
| ---- | ||
| % true | ||
| %% true | ||
| \%% false | ||
| %abc true | ||
| \%abc false |
There was a problem hiding this comment.
This would prevent a regression #13260 (comment)
| VALUES ('\'), ('\\'), ('\\\'), ('\\\\') | ||
| ---- | ||
| \ | ||
| \\ | ||
| \\\ | ||
| \\\\ |
There was a problem hiding this comment.
This query (and all other queries involving a \ backslash) would behave differently in the CLI
#13286
| @@ -829,3 +838,95 @@ SELECT | |||
| 'a' LIKE '%' | |||
| ---- | |||
| NULL true true | |||
|
|
|||
| # \ is an implicit escape character | |||
There was a problem hiding this comment.
When working on #13260 i didn't know that \ is an implicit escape character. I couldn't find any tests for this, so here they come.
| true true false true false | ||
|
|
||
| statement ok | ||
| create table inputs AS SELECT * FROM (VALUES ('%'), ('%%'), ('\%%'), ('%abc'), ('\%abc')) t(a); |
There was a problem hiding this comment.
Testing below query with VALUES should be enough... unless/until we have "push projection into values" optimization.
Maybe i should replace this with non-inlinable values using random?
There was a problem hiding this comment.
Testing below query with VALUES should be enough... unless/until we have "push projection into values" optimization.
I don't understand this -- there is only one column (t) and the query selects that column -- so there is no projection to push down 🤔
There was a problem hiding this comment.
it was for the case in which the inputs were not a table, but just VALUES ...
findepi
force-pushed
the
findepi/test-like-with-implicit-escape-51b346
branch
from
f33b075 to
86abe17
Compare
| ('%', '\%', '', ''), | ||
| ('_', '\_', '', ''), |
There was a problem hiding this comment.
This feels duplicate with what i added in string_literal, but adds test coverage for all string variants
findepi
force-pushed
the
findepi/test-like-with-implicit-escape-51b346
branch
2 times, most recently
from
d7de2e4 to
6c7d6ae
Compare
findepi
force-pushed
the
findepi/test-like-with-implicit-escape-51b346
branch
from
6c7d6ae to
aea67f1
Compare
findepi
force-pushed
the
findepi/test-like-with-implicit-escape-51b346
branch
from
aea67f1 to
6dd5744
Compare
| \a \\a false | ||
| \a % true | ||
| \a \% false | ||
| \a \\% false |
There was a problem hiding this comment.
This one looks like a bug.
Given \ is an escape character, \\ should match single backslash character \ of the input,
and % should match the rest (a). The result should be true then
| \\\\ \\a false | ||
| \\\\ % true | ||
| \\\\ \% false | ||
| \\\\ \\% false |
There was a problem hiding this comment.
This one looks like a bug.
Given \ is an escape character, \\ should match single backslash character \ of the input,
and % should match the rest (\\\). The result should be true then
|
@goldmedal added a few more test cases. Marked examples #13288 (comment) #13288 (comment) |
findepi
force-pushed
the
findepi/test-like-with-implicit-escape-51b346
branch
from
6dd5744 to
bc54c50
Compare
There was a problem hiding this comment.
Thank you @findepi and @goldmedal -- I think this does a great job documenting what our current LIKE behavior is. 🚀
| true true false true false | ||
|
|
||
| statement ok | ||
| create table inputs AS SELECT * FROM (VALUES ('%'), ('%%'), ('\%%'), ('%abc'), ('\%abc')) t(a); |
There was a problem hiding this comment.
Testing below query with VALUES should be enough... unless/until we have "push projection into values" optimization.
I don't understand this -- there is only one column (t) and the query selects that column -- so there is no projection to push down 🤔
More tests before #13260 lands.
Would prevent regression #13260 (comment)