* fix(sqlite): avoid using dash in fts5 queries
* test(db): get bookmarks and count with dash in keyword
* test(db): common and ad-hoc sqlite test case
After a first bookmark save, all subsequent ones fail because the update
query does not include an identifier and the update is done on all rows.
Introduced by 05fee53bd0
* chore: updated go-migrate dependencies
* fix: specify if we're saving bookmarks expecting a creation
up until now the SaveBookmarks method was doing some "magic" to do
"upserts" on the databases, but consistency between engines was scarce
and not knowing if we were expecting saving a new bookmark or updating
an existing one was leading to errors and inconsistencies in logic all
around the place. Now we need to specify a creation boolean when
saving and a differnt query will be make (INSERT vs UPDATE).
* fix(api): using incorrect bookmark for content downlaod
* test(db): added test pipeline for databases
Added functions that will share logic among the engines and will be
called on fresh databases on each test run
* dev: added basic docker-compose for development
* chore: uncommented tests
* ci(test): added mysql service
* typo
* test(mysql): select database after reset
* fix(mysql): ignore empty row errors when parsing tags
* fix(mysql): handle insert errors
* chore: added mysql variables to compose
* ci: explicit mysql service port exposed
* fix(psql): get last inserted id from insert query
book.ID was not being used, so inserts were failing.
the check for book.ID was removed and it is filled with the returning
id from the insert query
* test(psql): added save bookmarks simple test
* ci: added postgresql service
* fix(typo): QueryRow -> QueryRowContext
* ci: explicit postgresql port
* ci(test): 1.19 only
* ci: bind psql to localhost
* test(pg): migrate before test
* test(pg): migrate database before test
* fix(pg): check no rows error on get query
* [sqlite] Added has_content column and relevant migration
* Fix typo in docstring
* [sqlite] Fetch content for bookmarks separately
* [sqlite] Store has_content alongside with bookmark
* [sqlite] Rename variable to distinguish it from main query
* [sqlite] Use by-reference instead of no-op copy
* [sqlite] Reduce queries count from ≈30 to 1 to fetch tags
* Lint fixes
* Make minor logging improvements
* Wrap entire migration in transaction block
* Added «down» migration
* Drop workaround for old SQLite versions
It used an update and insert at the same time, since the virtual table
cannot have constrants nor upserts we need to try an update the
bookmark content to check for it's existence, inserting it only if the
update fails (don't affect any rows).
This will need to be improved in the future, since not all bookmark
updates require modification of the content table (in my case, I was
testing only updating tags).
