Set the number of enabled vertex attributes upon a shader program switch

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?

[monfera]

We ran into issues resulting in "glDrawArrays: attempt to render with no buffer attached to enabled attribute 1" and nothing rendered from that point on. Deriving from the standard, the way of avoiding the error seems to be either of:

1. managing which vertex attrib arrays are enabled and disabled
2. keeping vertex attrib arrays enabled but binding a mock or legacy buffer to it

The latter solution feels less robust in that it requires bindings to be left around 'just in case' a next shader program has fewer attributes and might benefit from the leftover binding.

This concept is the former solution. It works out fine with the plotly codebase on initial tests. There are two slight drawbacks: binding information to the gl object is safe but not pretty; use of introspection in render time. The latter problem is easily solved by flipping the comment line in the var newAttribCount lines.

I'd like to get feedback as this issue might have already come up and the resolution might have fallen some different way. As far as I know this approach is fairly common in tools that shuffle shaders in and out.

[mikolalysenko]

Thanks!

[etpinard]

@mikolalysenko would you mind publishing a new npm release with this patch?