Upgrade d3 formats

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.

[archmoj]

Attempt using latest d3-format and d3-time-format to resolve#5009 | Demo.

Notes:

• Need to figure out a why the test added in support locales in hovertemplate #3586 not passing.
• Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

@plotly/plotly_js

[nicolaskruchten]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

This seems like a showstopper.

[alexcjohnson]

Certain formats e.g. 0.f are no longer supported by d3 and throw an error.

There are two issues with this:

• We shouldn't ever throw an error, we should fall back on some default formatting and log an error (probably just once, otherwise this could easily dump thousands of logs)
• If possible we shouldn't lose any behavior that it's possible to keep.

Date formatting all goes through lib.formatDate - if there are cases where date formatting throws an error, we should be able to catch these there and convert them to errors. (@archmoj is that true? Any date formatting that we DON'T pass through lib.formatDate? If so, can we?)

Perhaps then we need a similar wrapper for number formatting to catch errors, and then we can use this wrapper to backfill removed functionality? It's not actually clear to me from the D3 changelog whether this is actually removed functionality, or perhaps it was previously falling back on a default but now throws an error instead?

If it's going to take a long time to be confident we're not breaking important behavior then sure, @nicolaskruchten's suggestion to upgrade only time is reasonable. But if it can be done quickly seems worth pressing ahead.

[archmoj]

Is it necessary to update d3-format as well as d3-time-format or can we just do the time update?

It seems necessary as they both depend on d3 locale.
So we'll need to make a similar wrapper for number formatting as suggested by Alex above.