Replaced Kross with QJSEngine and added example script to the test application.
Script code is compatible with Kexi 2 JS code
Removed the 'interpreter' property from a report as there is now only one interpreter
Test code will color alternate sections, set the caption of an item named 'label1' and print
a debug line on open and section render.
| No Linters Available |
| No Unit Test Coverage |
Attached image shows a script interacting with report objects (section and label). Code is visible in the output pane of QtCreator
| src/CMakeLists.txt | ||
|---|---|---|
| 6–7 | find_package(Qt5Qml REQUIRED) needed too | |
| src/items/check/KoReportScriptCheck.cpp | ||
| 20 | Preferred is #include <KProperty> for the current KProperty | |
| src/items/image/krscriptimage.cpp | ||
| 21 | staniekUnsubmitted Preferred is #include <KProperty> for the current KProperty | |
| src/renderer/scripting/krscripthandler.cpp | ||
| 32 | Convention is I guess that we don't use prefixes | |
| src/renderer/scripting/krscriptline.cpp | ||
| 21 | staniekUnsubmitted Preferred is #include <KProperty> for the current KProperty | |
One observation: if we want to integrate my SDC-based explicitly shared classes, please don't port all the elements. (to avoid unnecessary repeated work).
Despite of using Qt 5.5 I have not found better solution for making objects scriptable than using a wrapper QObject. That said, the branch scripting-T517-staniek of ssh://git@git.kde.org/clones/kreport/staniek/work shows that the wrapping is slightly different and SDC may be able to generate script methods for use (getters, setters).
Summing up we need to synchronize the efforts.
I conducted experiments with SDC also because of KDb APIs. For example DCount() function is so frequently used (also in reports) so we may want a globally consistent scripting approach.
There is no porting of elements required! :)
As of now, they work as in kexi2 with no changes to the wrapper classes. So you are happy to push this to master with above changes?
Fized all except #include <QtQml/....> as it the build doesnt find the headers otherwise
Fized all except #include <QtQml/....> as it the build doesnt find the headers otherwise
Do you have Qt5::Qml in target_link_libraries? Cmake will then add proper includes.
find_package(Qt5Qml REQUIRED) finds them, you need to have this line too.
Reviewed entire diff against kreport.git master. Comments below.
| src/common/KoReportData.cpp | ||
|---|---|---|
| 73 | Needed? | |
| src/renderer/KoReportPreRenderer.cpp | ||
| 509 | Don't we want to check result of trigger() and fail on errors? We're never sure if the report is valid if there's any error in scripts. | |
| 621 | So the owner is responsible for deleting the prev. version? | |
| src/renderer/scripting/krscripthandler.cpp | ||
| 31 | + please clean up the order of includes: separate local from Qt | |
| 91 | Check if code.isEmpty(), if so return immediately | |
| 91 | -> kreportDebug() | |
| 92 | Add m_reportData->script() as the 'const QString & fileName' arg. This can make the error message better. | |
| 98–99 | Do we want to call this also if m_scriptValue.isError()? | |
| 105 | Maybe it's better to destroy the engine first and then the exposed objects? | |
| 143 | Check result, display errors? | |
| 177 | Always true, isn't it? | |
| src/renderer/scripting/krscripthandler.h | ||
| 50–51 | change to bool trigger() | |
| src/renderer/scripting/krscriptreport.cpp | ||
| 103 | Is this property always present? | |
| 109 | Is this property always present? | |
| 115 | Is this property always present? | |
| src/renderer/scripting/krscriptreport.h | ||
| 46 | -> const QJSValue &val | |
| src/renderer/scripting/krscriptsection.cpp | ||
| 106 | Is this property always present? | |
| src/renderer/scripting/krscriptsection.h | ||
| 58 | -> const QJSValue &val | |
| src/wrtembed/KoReportDesigner.cpp | ||
| 288–289 | How exactly can we port it? Then set d->script but also remember value of report:script-interpreter. On saving save it back. Change the name obtained from report:script-interpreter to "JavaScript" as soon as scripts are edited so saving updates report:script-interpreter properly. | |
| 379–380 | Save the script only if it's not empty. | |
| 805 | Not needed | |
| 808 | Convention: use prepend() | |
| src/wrtembed/KoReportDesigner.cpp | ||
|---|---|---|
| 808 | OK... | |
Address comments by staniek 2
Need firther testing of loading/saving of interpreter and scripts
perhaps as further work to the example app using load of external report file
| src/CMakeLists.txt | ||
|---|---|---|
| 157 | Monir: Please keep order | |
| src/common/renderobjects.cpp | ||
| 49 | It's a c++11 construct. Asked on the list what's our policy for it in Calligra 3. | |
| src/renderer/KoReportScreenRenderer.cpp | ||
| 49 | just (!p) as usual | |
| src/renderer/scripting/krscripthandler.cpp | ||
| 114 | Why are these 3 deletes removed? | |
| src/wrtembed/KoReportDesigner.cpp | ||
| 200 | Not a native speaker but I propose | |
| 294 | Removed \n (better for the GUI) | |
| 297 | -> Unsupported Script Type | |
| 379–380 | What about this? | |
| src/renderer/scripting/krscripthandler.cpp | ||
|---|---|---|
| 114 | QJSEngine assumes ownership of the objects and deletes them when it is deleted ... took a while to figure out why double-free was happening!! http://doc.qt.io/qt-5/qjsengine.html#newQObject | |
| src/renderer/scripting/krscripthandler.cpp | ||
|---|---|---|
| 105 | Indeed! Thanks. | |
| src/CMakeLists.txt | ||
|---|---|---|
| 157 | was not needed anyway | |
Minor fix and go!
| src/wrtembed/KoReportDesigner.cpp | ||
|---|---|---|
| 297 | this is a title so All Caps needed -> "Unsupported Script Type" | |