My observation (on the current tip version 3.28.0) of Schrodingers Transactions is that if there is (for example) a transaction in progress and that is COMMIT or ROLLBACK, then the changes are either committed or rolled back and the explicit transaction is ended (that is, autocommit becomes True).
Statements which were in progress that were permitted to proceed (ie, where the next step did not return an abort error) continue with a read lock in place (ie, as if they were part of an implicit transaction on the connection) and once all those statements are completed, the read locks are released. You can BEGIN another transaction on the same connection (or another connection) and the locks will be escalated as you requested in the same fashion as would normally be expected for an in-progress implicit transaction. --- The fact that there's a Highway to Hell but only a Stairway to Heaven says a lot about anticipated traffic volume. >-----Original Message----- >From: sqlite-users [mailto:sqlite-users- >boun...@mailinglists.sqlite.org] On Behalf Of Keith Medcalf >Sent: Saturday, 2 March, 2019 19:52 >To: SQLite mailing list >Subject: Re: [sqlite] Handling ROLLBACK > > >Actually, in current versions of SQLite3 the error "(516) >SQLITE_ABORT_ROLLBACK" is returned by "other commands" and not by the >rollback command. If the ROLLBACK command completed successfully >(that is, it returned YaHoo! SQLITE_OK) then the transaction changes >have been rolled back. If "some other command" returns an error >indicating that it was ABORTed then that other command was aborted >and that has no bearing on whether or not changes to the database >were rolled back. They were. > >I presume that if a SELECT was in progress during the time that a >ROLLBACK was processed on the SAME CONNECTION and that the SELECT is >allowed to proceed, that it will maintain the REPEATABLE READ >isolation that it was commenced with and that when that statement is >finalized, the transaction context will be released. > >Whether the connection is still "inside" the transaction or not can >be determined by using the sqlite3_autocommit() API. > >Since this is kind of an undocumented grey area, I would suggest that >you do not make the programming errors which would result in this >situation (a Schrodinger Transaction state, for lack of a better >descriptive). > >--- >The fact that there's a Highway to Hell but only a Stairway to Heaven >says a lot about anticipated traffic volume. > > >>-----Original Message----- >>From: sqlite-users [mailto:sqlite-users- >>boun...@mailinglists.sqlite.org] On Behalf Of Keith Medcalf >>Sent: Saturday, 2 March, 2019 19:32 >>To: SQLite mailing list >>Subject: Re: [sqlite] Handling ROLLBACK >> >> >>>Suppose ROLLBACK does not cancel the BEGIN, can a programmer >>reliably >>>issue more SQL commands, including another ROLLBACK ? Will SQLite >>>continue to react correctly to other ROLLBACKs, and to SQL commands >>>which result in "(516) SQLITE_ABORT_ROLLBACK". >> >>A successful ROLLBACK on a transaction discards all changes that >were >>made during the transaction (rolls back to the state at the time the >>transaction began), and ends the transaction. It does not "cancel" >>the "BEGIN", it merely exits the transaction and discards changes >>made during the transaction. Since the BEGIN occurred in the past, >>it cannot be "cancelled" ... >> >>Full Stop End of Line. >> >>However, a ROLLBACK command which produces the error "(516) >>SQLITE_ABORT_ROLLBACK" would indicate that the ROLLBACK was NOT >>PERFORMED due to an error. Therefore the transaction is STILL IN >>EFFECT and has neither been COMMITed nor ROLLBACKed. Once the >>programming error which prevented the rollback is cleared the >>ROLLBACK command can be retried. Once a transaction is opened it >>remains in effect until ROLLBACK or COMMIT complete successfuly (or >>the heat death of the universe, whichever shall come first). >> >>Note that this discussion does not include "fatal errors" which may >>occur within the context of a transaction that abort the transaction >>itself. >> >>--- >>The fact that there's a Highway to Hell but only a Stairway to >Heaven >>says a lot about anticipated traffic volume. >> >> >> >>_______________________________________________ >>sqlite-users mailing list >>sqlite-users@mailinglists.sqlite.org >>http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users > > > >_______________________________________________ >sqlite-users mailing list >sqlite-users@mailinglists.sqlite.org >http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
