Initial support for accessing attached properties in-scope. by stephenmdangelo · Pull Request #22 · qmlweb/qmlweb-parser

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).

stephenmdangelo · 2016-08-09T18:09:46Z

This is to help support: qmlweb/qmlweb#297

Lots of tests fail, but I'm pretty sure that's just because all of the source is regenerated (though I believe still functionally equivalent). QmlWeb tests that fail are any that reference things like Qt and Imported, as there is no $QmlWebGetAttachedObject defined.

I'm mostly interested in any comments about this approach in general. It's currently including a copy of UglifyJS's process.js, with all of the node require/export stuff removed (I couldn't sort out a nice way to put that into replacements.js).

ChALkeR · 2016-08-11T23:35:07Z


+ function qml_get_attached_object(type, val) {
+ if (type === "name" && qml_is_element(val)) {
+ return "$QmlWebGetAttachedObject('" + val + "')";


I don't think this should be on the parser side. Parser should just provide a consumable AST.

ChALkeR · 2016-08-11T23:40:57Z

This adds a code generator to rebuild js functions code, correct?

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Could you also link me to an example why this is needed?

stephenmdangelo · 2016-08-15T18:32:02Z

This adds a code generator to rebuild js functions code, correct?

Yes, it takes the one from UglifyJS.

I am not sure this is the correct way (or even the repo) for that. I will need to think about this a bit. The initial plan was to have a codegen separately, but this in fact does relate to the parser, not a complete codegen.

Yeah, it's admittedly a hack (I don't even know if I can make this entirely transparent to completely arbitrary JS code that references capitalized variables). I'd love to hear some other ideas of how this could be done. I feel like we could make a babel plugin or something, but that would likely force the use of gulp-qmlweb, removing the ability to serve up QML files directly (or use Qt.createQmlObject, all the stuff being discussed in qmlweb/qmlweb#303).

Could you also link me to an example why this is needed?

Here is an example of why this is required: https://github.com/stephenmdangelo/playground/tree/master/qml/scoped_attached_properties

I'm less interested in being able to support qualified imports of QtQuick, but for custom attached properties (e.g. I'd like to add a custom attached property for QmlWeb that allows one to add CSS classes to an element. I've hacked that together for my own purposes via objectName).