```diff
// Filter is an interface to filter SQL
type Filter interface {
--- Do(sql string) string
+++ Do(ctx context.Context, sql string) string
}
```
### Adds a `Context` parameter to the `Do` method of the Filter interface.
Developers can rewrite SQL through the `Filter` `Do` method and **need to get the necessary data from the Context** to assist.
For example, get user information through `Context`, so that different users can use different tables. Another example is to get the flags through `Context` to add annotations to the SQL, and use the annotations to let the subsequent `DB-Proxy` to achieve read/write separation.
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2270
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: LinkinStars <linkinstar@foxmail.com>
Co-committed-by: LinkinStars <linkinstar@foxmail.com>
The image has already specified the entrypoint so we don't need to call `/cockroach/cockroach` in `cmd`.

Reviewed-on: https://gitea.com/xorm/xorm/pulls/2269
Co-authored-by: Zettat123 <zettat123@gmail.com>
Co-committed-by: Zettat123 <zettat123@gmail.com>
- Fixed quotes on drop table SQL statement.
- Implemented IsSequenceExists function for oracle dialect.
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2265
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Martin Viggiano <martinvigg@gmail.com>
Co-committed-by: Martin Viggiano <martinvigg@gmail.com>
This pr fixes a bug caused when a timestmap in TZ format with milliseconds (`2023-04-05T15:50:48.256816Z` or `2023-04-05T15:50:48.256816+08:00`) is being parsed.
- The bug is happens in the function `String2Time` in `convert/time.go`
- if the timestamp contains milliseconds a layout with the structure `2006-01-02 15:04:05.` is used to parse it
- This layout contains an space between the date and the hour, if the timestamp returned by the db contains a `T` the parser fails and an error is returned
This pr adds a check for the `T` and milliseconds in the timestamp, if both conditions are present the date is parsed using the RFC3339 layout
Added test cases to check timestamps in this format
Solves [#2244](https://gitea.com/xorm/xorm/issues/2244)
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2246
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Alberto Garcia <agarcia@circutor.com>
Co-committed-by: Alberto Garcia <agarcia@circutor.com>
This PR adds a `Truncate` method which allows to delete all existing rows in a table. The current `Delete` implementation enforces conditions to prevent accidental data deletion.
Co-authored-by: KN4CK3R <admin@oldschoolhack.me>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2220
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: KN4CK3R <kn4ck3r@noreply.gitea.io>
Co-committed-by: KN4CK3R <kn4ck3r@noreply.gitea.io>
When dropping indexes in Postgres if the index is non-regular we
should not attempt to regularise the index name as it is already correct.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2174
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Andrew Thornton <art27@cantab.net>
Co-committed-by: Andrew Thornton <art27@cantab.net>
`insertSelectPlaceholder` is an unexported const set at true. No code changes this nor can any build environment change it.
Therefore we should remove it and the associated dead code.
Close#2146
Signed-off-by: Andrew Thornton <art27@cantab.net>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2151
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Andrew Thornton <art27@cantab.net>
Co-committed-by: Andrew Thornton <art27@cantab.net>
The current mechanism for adding information about indices cannot express the ordering
of the columns in the index or add sorting information.
Here we add a new interface TableIndices which a struct would implement to provide
a slice of *schema.Index to provide additional indices to that gleaned from the tags.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2137
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Andrew Thornton <art27@cantab.net>
Co-committed-by: Andrew Thornton <art27@cantab.net>
MySQL/MariaDB: return max length for text columns using `CHARACTER_MAXIMUM_LENGTH`.
Tests:
* add an integration test: `TestGetColumnsLength`
* update `TestSyncTable3` since `TableInfo` isn't able to provide the column size
Co-authored-by: Pierre-Louis Bonicoli <pierre-louis.bonicoli@libregerbil.fr>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2133
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Pierre-Louis Bonicoli <pilou@noreply.gitea.io>
Co-committed-by: Pierre-Louis Bonicoli <pilou@noreply.gitea.io>
The [oldest unsupported version documentation](https://www.postgresql.org/docs/7.1/sql-comment.html) states that comment on a column is supported.
Update `TestGetColumnsComment` in order to check both MySQL/MariaDB and PostgreSQL.
Co-authored-by: Pierre-Louis Bonicoli <pierre-louis.bonicoli@libregerbil.fr>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2131
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Pierre-Louis Bonicoli <pilou@noreply.gitea.io>
Co-committed-by: Pierre-Louis Bonicoli <pilou@noreply.gitea.io>
ref pr https://gitea.com/xorm/xorm/pulls/2053
i think the previous fix has some issue
for example, i'm using session like this:
```go
// logSessionID == false
engine := NewEngine()
// use ctx.SessionId to distinguish uniq request id
cxt := context.WithValue(parent, log.SessionIDKey, "some unique request id")
session := engine.NewSession().Context(ctx)
```
however, with pr 2053, `session.Context` can't get SessionId from ctx.
this pr fix abrove issue, overwrite `session.Context()` only when `engine.logSessionID == true`
please check it out,thanks!
Co-authored-by: finelog <kaicltw@gmail.com>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2115
Co-authored-by: finelog <finelog@noreply.gitea.io>
Co-committed-by: finelog <finelog@noreply.gitea.io>
when using distinct cols with FindAndCount, reset statement.ColumnMap will make the counting sql an syntax error, this pr try fix this.
is this pr ok for merge?
error sql logging:
```sql
[SQL] SELECT DISTINCT `Fid` FROM `table_demo` WHERE Fkey = ? [val]
[SQL] SELECT count(DISTINCT ) FROM `table_demo` WHERE Fkey = ? [val]
```
after fix:
```sql
[SQL] SELECT DISTINCT `Fid` FROM `table_demo` WHERE Fkey = ? [val]
[SQL] SELECT count(DISTINCT `Fid`) FROM `table_demo` WHERE Fkey = ? [val]
```
Co-authored-by: finelog <kaicltw@gmail.com>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2096
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: finelog <finelog@noreply.gitea.io>
Co-committed-by: finelog <finelog@noreply.gitea.io>
dumpTables currently badly handles BLOB and TEXT data containing control
characters:
* MySQL will interpret and unescape string literals e.g.`\r` will become
carriage return.
* Postgres will not allow string literals to contain NUL nor will
SQLite so BLOBs will not dump correctly.
* Schemas should not be set on the destination dump
* MSSQL needs the N prefix to correctly ensure that UTF-8 data is
correctly transferred.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Co-authored-by: Lunny Xiao <xiaolunwen@gmail.com>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2091
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Andrew Thornton <art27@cantab.net>
Co-committed-by: Andrew Thornton <art27@cantab.net>
When dumping booleans these need to be converted from the original DB representation
to the new db representation. In the case of most DBs this is simply to 0 or 1 but
for postgres these have to be false or true.
Signed-off-by: Andrew Thornton <art27@cantab.net>
Reviewed-on: https://gitea.com/xorm/xorm/pulls/2089
Reviewed-by: Lunny Xiao <xiaolunwen@gmail.com>
Co-authored-by: Andrew Thornton <art27@cantab.net>
Co-committed-by: Andrew Thornton <art27@cantab.net>