overrideTags is highly memory inefficient

[zen0wu]

Based on our profiling result, it accounts for 5% of total memory allocation in our production servers.

Based on the implementation here,

hot-shots/lib/helpers.js

Lines 33 to 43 in f28d646

	formatTags(child, telegraf).forEach(tag => {
	const idx = typeof tag === 'string' ? tag.indexOf(':') : -1;
	if (idx < 1) { // Not found or first character
	toAppend.push(tag);
	} else {
	const key = tag.substring(0, idx);
	const value = tag.substring(idx + 1);
	childCopy[key] = childCopy[key] || [];
	childCopy[key].push(value);
	}
	});

It generates many intermediate strings by parsing the given formatted tags.

• When an array of tag pairs are parsed, [ "a:b", "c:d" ], it reparses them into { "a": "b", "c": "d" } and then reforms the strings
  • [ "a:b", "c:d" ] -> { "a": "b", "c": "d" } -> [ "a:b", "c:d" ]
• When an object is passed, it first format the tags using formatTags and then do the above. This is extra bad since it allocates "a:b", then take it apart, then reform "a:b". Might be the same memory wise due to v8's string interning, but at least it wastes many cpu cycles.
  • { "a": "b", "c": "d" } -> [ "a:b", "c:d" ] -> { "a": "b", "c": "d" } -> [ "a:b", "c:d" ]

[bdeitte]

Sorry for the slow response on this, and thanks for the info. As a heads up, no one is fixing any of the issues that come up here- this is all community-driven. So if you are interested in doing this, please feel free. Thanks.