Careful: 54cfa50 contains major changes & removes big-text hack
14 views
Skip to first unread message
Edward K. Ream
Jan 14, 2017, 7:03:27 AM1/14/17
to leo-editor
Please use extra caution for this rev.
Warning: This rev modifies the all-import tree-select logic. Problems here can (and have) caused data loss in the past. Be especially careful and report any problem immediately.
Having said this, the changes should be perfectly safe. They affect only syntax coloring, not the code that actually switches data. I have been eating my own dog food without problems. We may as well begin widespread testing now.
Here are the highlights, heavily modified from the checkin log:
A mass update, installing disabled pyzo coloring, and removing the big-text hack.
This is an interim fix for the extremely serious #273: Qt Bug involving large text nodes. The long-standing performance bug for legacy colorizing remains. Fixing it has high priority, and is required for closing #273.
It should not be possible to lose data in big text. Both the legacy colorizer (enabled by default) and the pyzo colorer cause Leo to wait until text has been fully loaded and colorized before displaying it. This is a major step forward, at the cost (sometimes) of big delays loading big text.
As an interim workaround for long colorizing delays, you can insert @killcolor as the first line of any node. This will disable the legacy colorizer and should provide acceptable performance.
The pyzo switch in leoColorizer.py controls whether the prototype pyzo syntax coloring is used. It is off by default. When the switch is on, everything is colored as if it were python, and the color uses hard-coded (solarized) styles.
Don't hold your work for improved pyzo coloring. There are many other higher-priority projects, and supporting 40+ languages with pyzo coloring is not one of them.
All questions and comments are welcome.
Edward
Warning: This rev modifies the all-import tree-select logic. Problems here can (and have) caused data loss in the past. Be especially careful and report any problem immediately.
Having said this, the changes should be perfectly safe. They affect only syntax coloring, not the code that actually switches data. I have been eating my own dog food without problems. We may as well begin widespread testing now.
Here are the highlights, heavily modified from the checkin log:
A mass update, installing disabled pyzo coloring, and removing the big-text hack.
This is an interim fix for the extremely serious #273: Qt Bug involving large text nodes. The long-standing performance bug for legacy colorizing remains. Fixing it has high priority, and is required for closing #273.
It should not be possible to lose data in big text. Both the legacy colorizer (enabled by default) and the pyzo colorer cause Leo to wait until text has been fully loaded and colorized before displaying it. This is a major step forward, at the cost (sometimes) of big delays loading big text.
As an interim workaround for long colorizing delays, you can insert @killcolor as the first line of any node. This will disable the legacy colorizer and should provide acceptable performance.
The pyzo switch in leoColorizer.py controls whether the prototype pyzo syntax coloring is used. It is off by default. When the switch is on, everything is colored as if it were python, and the color uses hard-coded (solarized) styles.
Don't hold your work for improved pyzo coloring. There are many other higher-priority projects, and supporting 40+ languages with pyzo coloring is not one of them.
All questions and comments are welcome.
Edward
Reply all
Reply to author
Forward
0 new messages