…e, query key-values. Data is currently held in a map.

… text files for Index and Data. Keys go to Index, whereas Values go to the Data file.

…, /?op=purge, /?op=compact. Repaired the overall logic of the server and disk persistence. More REST commands to: /quit, /config?trace_level=1111, /config?max_cache=1. Enhanced the demos/rest.cmd with a more complex workflow to demonstrate fragmentation and defragmentation of keys and values files.

… responses. Test batch rest.cmd exits after 100 iterations.

… table as well as any other user-table. The root of the web server is a generated HTML page with navigation for all the system features. There is now a map of tables. Selecting the current table is with the /tablename/?use rest command. Tables can also be dropped. Table::compact was reworked to shift-up deleted keys in the index. An index_revolver coroutine was introduced to return one index per invocation. This logic also has index cleanup capabilities. The Table::list() command uses this coroutine to produce the list of all keys. Table::get/set are now boolean. The set prevents overinserting and updating of unknown keys. The get returns false if the key is unknown. This allows now for inserting or updating "null" values. Better statistics. Better formatted output. All REST logic is in one DefaultHandle where the table list lock is maneaged. Everything the DB server does is listed in the / page which is built by DefaultHandle::build_help. Added the new Table::self_test method and the corresponding /?selftest=0 REST command. This system of self-testing obsoletes the rest.bat file while also aiming at raw performance.

This version has problems with deleting keys that are in cache. TODO.
This version closes issues: #1, #2, #3, #4, #5, #6, #8. It advances on issues #13 and #11.

… Closes issue: #14.

…elf_test.

…IO code should be reduced to this. Much better trace messages. Self tests. Corrections so that the self tests pass.

…s issue #16 regarding the encapsulation of IO operations. This version has a self-test as per issue #11.

…s issue #16 regarding the encapsulation of IO operations. This version has a self-test as per issue #11.

… to the main page before that. This is to avoid quit-restarting in a loop. Perparing to async-ify. A global and configurable mode_async. Added a rest provision to read the console log, if available. Reporting the sub-log of a Failed test when running in a loop. Git-ignoring Visual Studio temporary files and user-files of cmake. Working on issue #17.

…e back to the main page before that. This is to avoid quit-restarting in a loop. Perparing to async-ify. A global and configurable mode_async.

Added the static Table::co_self_test coroutine to wrap around run Table::self_test. The HTTP DefaultHandler calls either the self_test or the co_self_test depending on mode_async. In async mode, the self-test, which is a huge time penalty, is coroutined in the hope that it would split-up into individual, smaller, batches for the SHARD executor. This is to increase parallelism of the HTTP server when multiple users are concurrently active. Table::self_test is recursive in loop mode, and potentially endless. Merged the Prometheus sats server into the Database Demo server so only one port is exposed from the demo container. Added a rest provision to read the console log, if available. Reporting the sub-log of a Failed test when running in a loop. Git-ignoring Visual Studio temporary files and user-files of cmake. Working on issue #17.

…he index revolver, in favor of co_await table.co_index_locked() which is used by future Table::list and future Table::rownext.

…configurable sync/async table mode, as now there's just the async mode. Simplified get to return bool, internally use _get that returns iterator, which is also used by set. And it seems that I'm nolonger getting SIG35, so perhaps this closes #17.