Partial I/O when reading file from HTTP (S3)

[grzanka]

I was thinking of employing JSROOT as "online browser" of some large (>1GB) ntuples. I was planning to store the ntuples under publicly available HTTP address, on S3 storage (with CORS properly configured). The JSROOT could be part of static JS+HTML app served on GitHub pages.

Main aim would be to use "Open with player" option for ntuple. User could type the column names to plot, and possibly define cuts, like on example below:

I was curious how does I/O works when dealing with large ntuples. When I open the ntuple as URL what is being downloaded to my local browser session ? Only the columns which I select or whole file ?

I saw here:
#169 (comment)
that you mentioned partial file access.

[linev]

JSROOT uses partial I/O for reading files from http servers.

When performing TTree draw only data for selected branches read.
You can see actual HTTP requests in your browser. Just check this example:

https://root.cern/js/latest/?file=../files/hsimple.root&item=ntuple&opt=px

Very last HTTP request is reading baskets of px branch.
This simple example has only 4 baskets.

[grzanka]

JSROOT uses partial I/O for reading files from http servers.

Thanks ! I will test it .

[linev]

Thanks ! I will test it .

Be aware that some servers using http redirection - which sometime makes impossible usage of partial http requests

[grzanka]

With correct CORS settings it seems to work nicely with multiple GB files:

https://root.cern/js/latest/?file=https://s3p.cloud.cyfronet.pl/datarawlv2v3/20231204m2/64nA.root&item=channel_0/trc_file_000/segment_000/negative&opt=peak_area*0.07878780161263421:peak_value*0.07878780161263421%3E%3Eh(1000,-200,0,1000,-100,0)

The only piece which is puzzling me is the access to values in branches.
In the example above I've manually pasted 0.07878780161263421 constant.
It is however present in the channel_0/gain_mV leaf and saved as a string.
Do you know of any method to adjust my plotting formula and get that number directly from ntuple ?

[linev]

For the moment JSROOT do not allow to use data from other objects in TTree::Draw expression.
For such custom case you probably need to create your own HTML page and write JavaScript code there.

There is example of such code here:

https://root.cern/js/latest/api.htm#ttree_tselector

https://github.com/root-project/jsroot/blob/master/demo/read_tree.htm

[linev]

It is also possible directly use treeDraw() function of jsroot.
Here is example for node.js:

https://github.com/root-project/jsroot/blob/master/demo/node/tree_draw.js

Same code can be involved from the browser.
It is very easy to add several lines reading data from other objects in the same file.

[linev]

I closing issue.

Some better handling of TTree::Draw abort (when clicking on info message) included in https://github.com/root-project/jsroot/releases/tag/7.6.0