Chaining map triggers evaluation

[Dsolnik]

Description of expected behavior and the observed behavior

When chaining together multiple maps, we shouldn’t eagerly evaluate.

Complete, minimal, self-contained example code that reproduces the issue

In [1]: a = [1,2,3]

In [2]: def f(x):
   ...:     print(x)
   ...: 

In [3]: v = param.rx(a).rx.map(f)

In [4]: v = param.rx(a).rx.map(f).rx.map(f)
1
2
3

[jbednar]

That does seem undesirable. Care to try a PR to add a check to reactive_ops.map for an existing reactive op?

[camriddell]

Hi @jbednar I looked into this briefly, it seems that this is an intended effect of the code (when chaining a new rx, the previous rx is evaluated). Would love to take this on if OP isn't planning to and it's in-scope.

Here's another MRE based on an existing test:

items = []
def debug(value):
    items.append(value)
    return value

base_rx = rx('bob')
debug_rx = base_rx.rx.pipe(debug)
assert len(items) == 0

# this is False; the below `pipe` is evaluated because we chained another method
title_x = debug_rx.title()
assert len(items) == 0

# should evaluation be delayed until the value is explicitly accessed?
assert title_x.rx.value == 'Bob'
assert len(items) == 1

[philippjfr]

When I originally designed rx I wanted to have immediate validation but always intended to eventually add a completely lazy mode that defers evaluation until the value is actually used. Doesn't seem too difficult to add and then we can discuss which mode should be the default.