summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
...
* readme: clearly document which Zig version is supportedGravatar Vincent Rischmann2022-08-201-0/+8
|
* `addPackage` is redundantGravatar okamo2022-08-171-1/+1
|
* change `[]const u8` to `std.build.FileSource`Gravatar okamo2022-08-171-1/+1
|
* change a way to use bundled sqlite3Gravatar okamo2022-08-171-1/+1
|
* make `c` public againGravatar Vincent Rischmann2022-08-151-1/+1
| | | | Accidentally made it private during a refactoring
* require SQLite >= 3.21.0Gravatar Vincent Rischmann2022-08-143-7/+13
| | | | | | | | | | | | After some testing it turns out we need at least SQLite 3.21.0 to compile. We _could_ work around this and make zig-sqlite work with older SQLite versions but I don't think it's necessary because: * "old" distributions like Debian Buster, RHEL 8 ship with SQLite > 3.21.0 * in any case if people want to build for OSes where SQLite is too old they can use the bundled source code.
* ci: only run the workflow on a push to masterGravatar Vincent Rischmann2022-08-061-0/+1
| | | | | | | | With both `push` and `pull_request` a push to an open PR triggers two redundant builds which we should avoid. This should have minimal impact since I typically open a PR when working on a feature.
* ci: cache the zig cache filesGravatar Vincent Rischmann2022-08-061-0/+8
|
* ci: cancel in progress jobGravatar Vincent Rischmann2022-08-061-0/+4
|
* ci: run the workflow on all changesGravatar Vincent Rischmann2022-08-061-8/+0
|
* move versionGreaterThanOrEqualTo in sqlite.zigGravatar Vincent Rischmann2022-08-052-4/+7
| | | | We will need it later on.
* move Text in sqlite.zigGravatar Vincent Rischmann2022-08-052-4/+4
| | | | | | | | | At first Blob and Text were defined in query.zig because they were used for bind parameters, but now we also use them for reading text or blob columns appropriately. Both types now live in sqlite.zig which query.zig already imports so it doesn't change anything.
* move DetailedError, do some cleanupGravatar Vincent Rischmann2022-08-042-49/+53
| | | | | | | | DetailedError uses getLastErrorOffset, which must check the SQLite version with versionGreaterThanOrEqualTo. Instead of publicly exposing versionGreaterThanOrEqualTo in sqlite.zig move DetailedError in errors.zig
* use a single cImportGravatar Vincent Rischmann2022-08-043-6/+5
| | | | | | | | | Two cImport calls generate incompatible code: we can't use the structs generated in sqlite.zig in functions in errors.zig for example Up until now it wasn't actually a problem since in errors.zig we only ever used constants which does work, but now we want to introduce functions in this file.
* add greaterThanOrEqualsToGravatar jiacai20502022-08-041-5/+9
|
* ci: also trigger on ref creationGravatar Vincent Rischmann2022-08-021-0/+1
|
* ci: test on macOS 12Gravatar Vincent Rischmann2022-08-021-1/+1
|
* ci: test on all Ubuntu's variantsGravatar Vincent Rischmann2022-08-021-1/+1
| | | | | | | Ubuntu 18.04 actually uses SQLite 3.22.x and as seen in 530feae7884ae5942c50cba30a307ddbcd767402 the SQLITE_DIRECTONLY flag doesn't exist until SQLite 3.30.0. If we had tested with Ubuntu 18.04 we would have caught this earlier.
* remove obsolete todoGravatar Vincent Rischmann2022-08-021-3/+0
|
* no need for 'comptime' for this checkGravatar Vincent Rischmann2022-08-021-1/+1
| | | | A comparison against a const is already comptime.
* create a specific CreateFunctionFlag struct based on the SQLite versionGravatar Vincent Rischmann2022-08-021-2/+16
|
* some error codes are only available with sqlite >= 3.22.0Gravatar Vincent Rischmann2022-08-021-5/+11
| | | | | | | We wrongly assumed they were always present but after testing with SQLite 3.21.0 it's not the case. I think they were missed because their first version is not documented.
* fix buildGravatar Vincent Rischmann2022-07-261-1/+1
|
* build: let the user specify the build modeGravatar Vincent Rischmann2022-07-231-8/+8
|
* add a way to get the aggregate context with createAggregateFunctionGravatar Vincent Rischmann2022-07-141-53/+161
| | | | | | | | | | | | | | | | | | | | The old way of working was that we always passed the user context as first argument to both `step` and `finalize` functions and the caller had no way of getting the aggregate context from SQLite (http://www3.sqlite.org/c3ref/aggregate_context.html). Now both `step` and `finalize` functions must have a first argument of type `FunctionContext`: fn step(fctx: FunctionContext, input: u32) void { var ctx = fctx.aggregateContext(*u32) orelse return; ctx.* += input; } fn finalize(ctx: *u32) u32 { var ctx = fctx.aggregateContext(*u32) orelse return 0; return ctx.sum; } Fixes #89
* readme: fix allocator usageGravatar Vincent Rischmann2022-06-261-1/+1
|
* Merge branch 'clarify-bind-markers-error'Gravatar Vincent Rischmann2022-05-241-1/+1
|\
| * clarify the compile errorGravatar Vincent Rischmann2022-05-241-1/+1
|/
* Make ParsedQuery a generic on query lengthGravatar Luna2022-05-172-163/+170
|
* Make errorFromResultCode publicGravatar luna2022-05-151-0/+1
| | | Useful for direct C API users.
* remove unreachable conditionGravatar Luna2022-05-141-2/+0
|
* add test for single statement in execMultiGravatar Luna2022-05-141-1/+9
|
* rename to execMultiGravatar Luna2022-05-141-2/+2
|
* add newlines and comments to runMulti test caseGravatar Luna2022-05-141-1/+1
|
* fix splitting logic for runMultiGravatar Luna2022-05-141-4/+21
|
* copy options overGravatar Luna2022-05-141-4/+5
|
* add Db.runMultiGravatar Luna2022-05-141-1/+24
|
* add a test for a bind marker with an optional typeGravatar Vincent Rischmann2022-05-031-14/+34
|
* allow parsing optional types in bind marker typesGravatar Vincent Rischmann2022-05-031-1/+16
|
* Merge branch 'fix-bind-identifier-parsing'Gravatar Vincent Rischmann2022-05-022-3/+8
|\
| * the '_' character is valid in a named bind parameterGravatar Vincent Rischmann2022-05-022-3/+8
|/
* put the TODO comment at the end of the line insteadGravatar Vincent Rischmann2022-05-021-9/+3
| | | | | | | | | | When using an explicit error set one must add the "Workaround" error in their error set; putting the explanation as to why it exists at the end of the line makes it so it is displayed directly in the output of the compiler. Before a user had to go look at the source code to understand why we have this workaround.
* Makes sqlite.Error pub.Gravatar Felix "xq" Queißner2022-04-291-1/+2
|
* remove unused assertGravatar Vincent Rischmann2022-04-241-3/+0
|
* clarify type nameGravatar Vincent Rischmann2022-04-241-3/+3
|
* add compile error when passing non-struct to Statement.bindGravatar Luna2022-04-241-0/+5
|
* add test for runtime slices as DynamicStatement argsGravatar Luna2022-04-241-0/+22
|
* emit compileError on unsupported pointer sizesGravatar Luna2022-04-241-4/+8
|
* add support for arrays as runtime bind valuesGravatar Luna2022-04-241-0/+5
|
* allow slices to be passed as bind parametersGravatar Luna2022-04-241-9/+22
|
HIC SVNT DRACONES