Out of curiosity I did a speed comparison for various ways to access data in DataTiddlers. This is the result as measured with Maurycy’s Advanced Performance Plugin:
| Filter | Total time | Uses | Longest run | Average time | Median time |
|---|---|---|---|---|---|
| [<jsondata>jsonget[807810887407]] | 6942.00ms | 1000 | 17.00ms | 6.94ms | 6.00ms |
| [{New Tiddler 4}jsonget[807810887407]] | 6935.00ms | 1000 | 27.00ms | 6.93ms | 6.00ms |
| [[New Tiddler 3]getindex[807810887407]] | 62.00ms | 1000 | 1.00ms | 0.06ms | 0.00ms |
| [[New Tiddler 4]getindex[807810887407]] | 15.00ms | 1000 | 1.00ms | 0.01ms | 0.00ms |
jsonget[] is comparatively slow for simple name:value DataTiddlers. The faster getindex[] is 4× slower for application/x-tiddler-dictionary than for application/json, both being in the microseconds per lookup.
Here’s how it was measured:
Tiddler New Tiddler 3 is a dictionary tiddler with 18075 entries of random numbers. The last lines look like this:
[...]
133007463997:142745435157
176656710595:430569764937
133014656883:396363251674
474946751459:374715567685
807810887407:631723026204
Tiddler New Tiddler 4 is a JSON Tiddler with the same data in JSON format. The last lines look like this:
[...]
"133007463997": "142745435157",
"176656710595": "430569764937",
"133014656883": "396363251674",
"474946751459": "374715567685",
"807810887407": "631723026204"
}
In another tiddler I ran a lookup of the last data point (with the name 807810887407) 1000 times:
<$let jsondata={{New Tiddler 4}}>
<$list filter="[range[1000]]">
<$let a={{{ [[New Tiddler 3]getindex[807810887407]] }}}
b={{{ [[New Tiddler 4]getindex[807810887407]] }}}
c={{{ [{New Tiddler 4}jsonget[807810887407]] }}}
d={{{ [<jsondata>jsonget[807810887407]] }}}>
</$let>
</$list>
</$let>
and measured its render time.
Just in case that the elapsed time for jsonget[] was due to loading the text data from the JSON tiddler, I also assigned it to a variable and read it from there. There was not much of a difference, though, see the table above.
I have no idea if the getindex[] lookups are cached and therefore faster when always looking up the same key.
Also, I explicitly don’t want to make jsonget[] look bad or anything – it’s much more powerful than getindex[] after all. For simple data tiddlers, I’ll stick with getindex[]. Not that I have many of those left, since Jeremy wants to deprecate them and all…
Have a nice day
Yaisog