On Jan 13, Brian Vogel wrote:

The regex processing as currently implemented by default also makes ([A-Z]) actually capture ([A-Z,a-z]) or ([A-z]), so it's no surprise that I had
gibberish, as it is capturing all kinds of letters I don't intend it to unless I turn case sensitivity on.Just plain dumb in the regex world.

True, but as pointed out: every tool in the Unix world (where these were developed), that does regex matching, provides an "ignore case" switch that will do this same thing.

Though it's never by default, and I agree that it is not what anyone using character classes would expect to be the default, unless they notice the checkbox.

If you put in a feature request to have a warning pop-up when using regular expressions and case sensitivity is off, I will implement the feature when I get the chance, assuming the devs don't object.

The main objection will likely be: anyone using regular expressions is technically savvy enough to pay attention to how their checkboxes are set.
If you do put in a feature request, I suggest you address that up front.

Luke

The regex processing as currently implemented by default also makes
([A-Z]) actually capture ([A-Z,a-z]) or ([A-z]), so it's no surprise that

…I am very grudgingly ‘tech savvy enough’ to deal with regexes.  Barely.  Because they break my brain and the only reason I learned them was to be able to coerce Nvda into telling me if a word is capitalized when reading text.  

That’s why I posses most of my tech skills, to be honest- I can’t force the software to do what I need it to do until I pick up all sorts of low level skills like regexs. 

The less unexpected brain breaking things I have to learn, the better- and a wierd checkbox that is sort of hanging out in the middle of the dialog, just begging to be skipped by a grumpy user in a hurry to make things work…that’s why I would check that box by default, because the first regrx I made captured EVERY letter and not just the capitals and barred out gibberish. 

And you can tell when I wrote my ‘here’s my janky hack to speak proper nouns out loud’ post, I didn’t even remember to add the ‘you need to check ‘case sensitive’ box to make this work’ because it’s the sort of detail that slipped my mind.  It’s a very mind-slipping detail.

For reference: 

https://nvda.groups.io/g/nvda/message/102408

-Tara

On Sat, Jan 14, 2023 at 1:21 AM Luke Davis <luke@...> wrote:

On Jan 13, Brian Vogel wrote:

> The regex processing as currently implemented by default also makes ([A-Z]) actually capture ([A-Z,a-z]) or ([A-z]), so it's no surprise that I had
> gibberish, as it is capturing all kinds of letters I don't intend it to unless I turn case sensitivity on.Just plain dumb in the regex world.

True, but as pointed out: every tool in the Unix world (where these were
developed), that does regex matching, provides an "ignore case" switch that will
do this same thing.

Though it's never by default, and I agree that it is not what anyone using
character classes would expect to be the default, unless they notice the
checkbox.

If you put in a feature request to have a warning pop-up when using regular
expressions and case sensitivity is off, I will implement the feature when I get the chance,
assuming the devs don't object.

The main objection will likely be: anyone using regular expressions is
technically savvy enough to pay attention to how their checkboxes are set.
If you do put in a feature request, I suggest you address that up front.

Luke

Tyler,

Using your example of testing, which I should have done, I am using this sentence as the test sentence:
I have no idea why there is No one who can say NO and stick to it.

The first no is lowercase, second capital N lowercase O, and last uppercase.  My expectation, prior to any testing using the regex:
[Nn]o
as the matching pattern with 
Yes
being the replacement, that what NVDA should say when reading that sentence would be: I have yes idea why there is yes one who can say NO and stick to it.
Checking the "case sensitive" checkbox should have zero effect, as this should not be applied when a regex pattern is used.

But when I do use exactly that regex, all three "no"s in that sentence get matched when case sensitivity is NOT checked.  Yet the regex itself should NOT match an all uppercase NO.

This is freakin' confusing to someone who has a very long history indeed with regular expressions.  A regular expression pattern match should use precisely what is specified by that regular expression, nothing more, nothing less.  If I had wanted to capture and substitute all three cases the regex would have been [Nn][oO], and that should capture nO (lowercase N with uppercase O), too, but the original one should not.
--

Brian - Virginia, USA - Windows 11 Pro, 64-Bit, Version 22H2, Build 22621; Office 2016, Version 16.0.15726.20188, 32-bit

It is much easier to be critical than to be correct.

       ~ Benjamin Disraeli, 1804-1881

In all my years using Regular Expressions, I never use any case sensitivity indicatior such as the checkbox when creating NVDA dictionary entries, because the structure of the regex should be handling that itself.

But, I noticed in the SayCap.dic file its creator has case sensitivity turned on, even though the regular expressions do handle case sensitivity as written.

I don't know anything about how NVDA may handle a checked case sensitive box with a regular expression match.  I would think (hope, fervently) that it would be ignored if checked for a regular expression match, but I have learned the hard way that very peculiar things can happen at times with the NVDA dictionaries.

I suspect that the fact that the dictionary processing does not drop out when a match and substitution is initially made, but keeps passing the replacement string "down the line" to be checked against subsequent entries, may be causing some of my "gibberish" issues.  But I have no idea how/whether/if the case sensitive option could have any impact whatsoever in processing if a regex match is used.
--

Brian - Virginia, USA - Windows 11 Pro, 64-Bit, Version 22H2, Build 22621; Office 2016, Version 16.0.15726.20188, 32-bit

It is much easier to be critical than to be correct.

       ~ Benjamin Disraeli, 1804-1881

In all my years using Regular Expressions, I never use any case sensitivity indicatior such as the checkbox when creating NVDA dictionary entries, because the structure of the regex should be handling that itself.
But, I noticed in the SayCap.dic file its creator has case sensitivity turned on, even though the regular expressions do handle case sensitivity as written.
I don't know anything about how NVDA may handle a checked case sensitive box with a regular expression match.  I would think (hope, fervently) that it would be ignored if checked for a regular expression match, but I have learned the hard way that very peculiar things can happen at times with the NVDA dictionaries.
I suspect that the fact that the dictionary processing does not drop out when a match and substitution is initially made, but keeps passing the replacement string "down the line" to be checked against subsequent entries, may be causing some of my "gibberish" issues.  But I have no idea how/whether/if the case sensitive option could have any impact whatsoever in processing if a regex match is used.
--
Brian - Virginia, USA- Windows 11 Pro, 64-Bit, Version 22H2, Build 22621; Office 2016, Version 16.0.15726.20188, 32-bit
*/It is much easier to be critical than to be correct./*
       ~ Benjamin Disraeli, 1804-1881