On Thu, Mar 03 2022, Jaydeep Das wrote:

How about modifying the number match regex to:

`[0-9._]+([Ee][-+]?[0-9]+)?[fFlLuU]*[^a-zA-Z]` ?

The `[^a-zA-Z]` in the end would make sure to not match
the `.F` in `X.Find`.

Additionally, we can add another regex for matching just
the method calls:

`[.][a-zA-Z()0-9]+`

Both of these changes would make word_regex match 2 tokens in
X.Find() : X and .Find() (Here X can be any valid identifier name)

quoted

How many tokens will the word-regex find in the expression X.e+200UL?
.e+200UL is a single token. > It's most easily fixed by requiring a digit before the fullstop. But if
floatingpoint numbers can begin with a fullstop, then we need a second
expression that requires a digit after a leading fullstop.

But that syntax would be wrong. I tried making a condition like you said,
but it always ended up breaking something else(like breaking 2.e+200UL into 2, .e, + and 200UL)

Also, I realized I did a bit of mistake in the identifier regex.
Both _abc and __abc are valid identifiers. _3432, __3232 are valid identifiers too.(not numbers)

The previous regex matched only one `_`, so in the next patch,
I plan to implement the following regex:

Identifier: `([_]*[a-zA-Z]|[_]+[0-9]+)[a-zA-Z0-9_]*`

Numbers: `[0-9_.]+([Ee][-+]?[0-9]+)?[fFlLuU]*[^a-zA-Z]`
(It makes sure that in X.Find, .F is not matched )

Additionally, An extra regex for method calls:

`[.][a-zA-Z()0-9]+`

What do you think?

Just a small note on rx syntax> [.] can be handy to escape "." (but you
can also use "\\.", but that's arguably not as easy to read.

But there's no reason to use [_]* over just _*..

(Also, I have an in-flight change to userdiff.c that would conflict, but
I wonder if it wouldn't be handy to make the word_regex a "struct
userdiff_funcname". Then we could specify icase flags, which in this
case would make it a lot easier to read).