Not logged in

Many hyperlinks are disabled.
Use anonymous login to enable hyperlinks.

The initial check-in for each branch:

20:22 [1122ff55e5] Closed-Leaf: Combine the OP_MakeIdxKey and OP_MakeKey opcodes into a single OP_MakeKey that does the work of both. Standardize the parameter meanings on OP_MakeRecord and OP_MakeKey. Remove the unused OP_Seek opcode. Other related code simplifications and cleanup. (user: drh, tags: refactor-MakeRecord)
14:16 [3fe22bbdf0] Closed-Leaf: Checked in untested changes by mistake. (user: drh, tags: mistake)
20:01 [bc9c9f73c5] Updates to use the next-generation-query-planner from the SQLite3 project. This branch is largely broken. (user: dan, tags: nextgen-query-planner)
20:28 [d94f6e934e] Start using sqlite4_num to store numeric SQL values. This commit is more buggy than not. (user: dan, tags: sqlite4-num)
20:15 [abe0420724] Allow accessing a prefix of the database file using mmap and the remainder using read and write. (user: dan, tags: prefix-mmap)
14:18 [35af051367] Leaf: Add dedicated get/set functions for the schema cookie to the sqlite4_kv_methods structure. (user: dan, tags: experimental)
19:42 [69f33cfa12] Add definitions for the extra locks required for read-only clients to detect whether or not a database is live. (user: dan, tags: read-only-clients)
05:27 [9ff6991d9f] Make LARGEST_UINT64 unsigned. This was causing adjustExponent to fail when compiled with msvc. (user: peterreid, tags: num_work)
19:03 [3bf1db9709] Add a field to the database header to identify the compression scheme in use. (user: dan, tags: compression-id)
19:04 [d1b1a9e969] Map and unmap parts of the database file on an LRU basis to limit the amount of address space consumed at any one time (for 32-bit address spaces). It looks like this might be slower than read() and write() anyway... (user: dan, tags: mmap-on-demand)
20:07 [09251cee6a] Change the lsm file-format to allow a small number of blocks belonging to the oldest segment in the database to be moved without modifying their content. This makes it easier to compact a database file to the minimum required size. (user: dan, tags: block-redirects)
19:56 [a235305d42] Add APIs to allow fts5 to be augmented with ranking and snippet functions. Does not work yet. (user: dan, tags: matchinfo)
20:13 [9374c3a283] Change free-list deletes to use range-deletes instead of point-deletes (so that they can be coalesced when segments are merged). This has revealed problems with the range-delete code. (user: dan, tags: range-delete-fixes)
20:04 [ebca1063ac] Rework the free block list storage so that it scales properly. Currently some test cases fail. (user: dan, tags: freelist-rework)
15:26 [5d266a717d] Change page numbers to 8-byte numbers (from 4). This is required to support compressed databases, where a page number is a byte offset in the database file. (user: dan, tags: compression-hooks)
20:05 [daa6942834] Add (partially working) code for deleting keys to lsm_tree.c. Required for range-deletes. (user: dan, tags: range-delete)
19:38 [50f8b55823] Rework flow control some (flow control = slowing down clients when worker threads or processes cannot keep up). (user: dan, tags: rework-flow-control)
21:58 [6d31a9d20c] Create new branch named "lsm_env-xsize" (user: stephan, tags: lsm_env-xsize)
17:42 [daac1c7c4e] Add file "shm.wiki" with notes regarding shared memory contents. (user: dan, tags: multi-process)
22:42 [3abadc2e11] Closed-Leaf: "uplifted" sqlite4::init to Sqlite4InitInfo typedef. (user: stephan, tags: stephan)
20:29 [1d65209131] Continuing work on sqlite4_env and getting it to be used everywhere. The changes here are mostly having to do with mutexes. They compile, but there are errors. This is an incremental check-in so that I can change to work on another project where an issue has just come up. (user: drh, tags: sqlite4_env)
02:37 [196691d50e] Begin refactoring the memory allocation logic to always require an environment pointer. This code will compile but does not run. (user: drh, tags: malloc-refactor)
20:22 [97fe6585f1] Experimentally try embedding b-tree pages in sorted runs instead of creating an external separators run. This slows down the rate at which blocks can be cycled during incremental merging, but also makes database writes more contiguous. (user: dan, tags: embedded-btree)
19:52 [3841829752] Changes to support "real" user-defined primary keys. This is quite broken at present. (user: dan, tags: primary-keys)
21:47 [87d659d624] This is an incremental check-in for the massive change that removes the old b-tree layer and substitutes a uniform key-value namespace. Nothing works at this point. This check-in is just a checkpoint. (user: drh, tags: remove-btree)
01:18 [84d5dea8fd] initial empty check-in (user: drh, tags: trunk)