| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
Resolves a typo (dervied->derived) in sqlite prepared query var, which was preventing the derivedcompounddef table from populating.
Fixing this issue also exposed an underlying problem where junk data was showing up in two columns. Porting some if/else logic from the code that inserts to basecompounddef seems to have fixed this, as well.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The uniqueness of rows is maintained by first doing a select.
In case the item is not found, it's INSERTed.
This is done for files, refids, params and includes, data
that is not unique by itself, like cross-references for example.
In other words, you can have many memberdefs in one file,
but you can only have one cross-reference between two functions
since the reference location is unique.
So, when SELECTing files and refids, use the file name and refids
as primary key since that's the fastest way to do it.
Also 'OR REPLACE' not necessary in 'INSERT OR REPLACE'.
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* template and function arguments
* innernamespaces
* put sqlite3_stmt and the coresponding query in
a struct
* use has(Multi|One)LineInitializer instead of hardcoded value
when extracting references from initializers
* sqlite3_clear_bindings in step()
* getRow in step() should use different sqlite3_ functions
depending on wether is doing 'INSERT' or 'SELECT'
includes table uses only file IDs
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
|
| |
In table basecompoundref, use SQLITE_TRANSIENT for
base, derived, refid
|
|
|
|
|
|
|
|
|
|
| |
memberdef table changes:
* rename bstart to bodystart
* rename id_bfile to id_bodyfile
* add bodyend
* add {brief,detailed,inbody}description
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
|
|
|
| |
* call sqlite3_clear_bindings when step fails
* more miss-placed returns
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
|
|
| |
these were causing a major performance hit
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
|
|
|
| |
the code comes from the early stage of sqlite3gen, when xref keys were
kept as "file:line:column", not as function name.
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
|
|
|
|
|
|
| |
return-ing on a "step" failure, drops
any info that could be extracted further on.
instead, call sqlite3_clear_bindings()
statements fail and carry on
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
|
|
| |
this way, it's easier to see what's missing from sqlite3gen
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <groleo@gmail.com>
|
| |
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
| |
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
|
|
|
| |
Signed-off-by: Adrian Negreanu <adrian.m.negreanu@intel.com>
|
| |
|
| |
|
|
|