kicad-developers team mailing list archive

[Henner Zeller <h.zeller@xxxxxxx>]

On 18 December 2015 at 21:57, Henner Zeller <h.zeller@xxxxxxx> wrote:
> On 18 December 2015 at 21:50, Chris Pavlina <pavlina.chris@xxxxxxxxx> wrote:
>> On Fri, Dec 18, 2015 at 09:40:29PM -0800, Henner Zeller wrote:
>>> On 18 December 2015 at 21:33, Chris Pavlina <pavlina.chris@xxxxxxxxx> wrote:
>>> > I'm not going to argue about search methods, it'll suffice to say I
>>> > disagree. I don't _want_ to search my library your way, and yes, I
>>> > understand that it can work.
>>> >
>>> > I'm unconvinced that adding one option box is excessive complexity. It's
>>> > not in the main UI, it's in the preferences dialog. What's wrong with
>>> > preferences? Users can ignore those. Most do. Though the preferences
>>> > dialog does need to be reorganized.
>>>
>>> The problem is, that people don't find it. They get frustrated because
>>> they want to search with regular expressions, but it doesn't work. And
>>> they never find the option to set it.
>>
>> Somehow I don't think that wildcard/regex search is going to be a big
>> advertised feature that all the kids will be flocking to kicad for. I
>> also doubt that the sort of person who would want it is the sort of
>> person who couldn't be bothered to look at the options.
>>
>>>
>>> If the dialog would do all matches in parallel and increases scores
>>> for each matcher that triggers, then you get all the simple matching I
>>> am proposing and all the extended regular expression searching that
>>> you want. And it will work automatically without anyone ever setting
>>> an option.
>>
>> Yes, and clutters up the results with false matches, just like searching
>> for "ATXMEGA*D3" by typing "ATXMEGA D3" does.
>
> But if you have all scoring functions engaged, the relevant come first.
>
>>
>> Seriously, what is it with this mailing list? Every time someone
>> suggests a simple idea, everyone immediately piles on with their
>> favorite way to overcomplicate it.
>
> Calm down. I make suggestions to _simplify_, not complicate.
>
>>
>> I just suggested adding this because I thought it'd be a simple addition
>> that might add some utility for some people. Never imagined wildcards
>> could be so controversial.
>
> As I said, I like wildcards and regexp and like to see them supported.
> But I suggested way to support them automatically.
>
>>Take it or leave it, I'm not going to keep
>> revising it to make it more complex.
>
> You don't have to. I can do that.

Attached, the simplified patch, using your new matchers:

   * Uses all your matchers, but does not need to have the factory.
   * No need for extra code to add option for users.
   * Same effect. People can use any syntax they want and get matches
accordingly.
   * It is very unlikely to get unwanted results.
   * Less code (366 vs 833 loc)
   * More user friendly to use (no search for obscure options, if
someone uses FOO*BAR or a regexp it just works as expected).

-h

>
>>
>>>
>>> -h
>>>
>>> >
>>> >
>>> > On Fri, Dec 18, 2015 at 09:19:54PM -0800, Henner Zeller wrote:
>>> >> On 18 December 2015 at 20:16, Chris Pavlina <pavlina.chris@xxxxxxxxx> wrote:
>>> >> > As discussed earlier today, this patch adds support for both wildcard
>>> >> > and regular expression search to the eeschema component chooser. An
>>> >> > option is added to eeschema options to select the "Component search
>>> >> > method", which defaults to "Simple" (the original behavior). Regular
>>> >> > expression search was added as a "bonus", as it was used as a stepping
>>> >> > stone to wildcard search.
>>> >>
>>> >> Cool, I'll try it.
>>> >>
>>> >> BUT: I am skeptical I must admit, as adding these features that are
>>> >> undoublty more complicated for the non-software engineer to use and
>>> >> also to choose from. More choices for the user to handle, and I am not
>>> >> convinced yet that it is worth it.
>>> >>
>>> >> I would rather try to make the search as best as possible that a few
>>> >> words will bring the result. If it doesn't, then the scoring function
>>> >> needs to be tweaked. I am a big fan of simple user interfaces...
>>> >>
>>> >> Do you have some real world examples that are really hard to solve
>>> >> with simple string-matching and scoring of the result as it is now
>>> >> (the XMEGA example in the other thread can be done by just having a
>>> >> space between the xmega and d3, so this is not really a good example).
>>> >>
>>> >> Having simple user interfaces is hugely important IMHO, so we should
>>> >> add features and confusing user options only if the simple way of
>>> >> dealing with it can't do it.
>>> >>
>>> >> -h
>>> >>
>>> >> >
>>> >> > The existing behavior of Henner's component chooser wasn't changed, just
>>> >> > the matching function it uses.
>>> >> >
>>> >> > Following Wayne's suggestion, an EDA_PATTERN_MATCH base class is defined
>>> >> > in include/eda_pattern_match.h, and then:
>>> >> >
>>> >> > - EDA_PATTERN_MATCH_SUBSTR implements EDA_PATTERN_MATCH with the old
>>> >> >   behavior
>>> >> >
>>> >> > - EDA_PATTERN_MATCH_REGEX implements EDA_PATTERN_MATCH providing regex
>>> >> >   search via wxRegEx and wxRE_ADVANCED.
>>> >> >
>>> >> > - EDA_PATTERN_MATCH_WILDCARD extends EDA_PATTERN_MATCH_REGEX,
>>> >> >   translating a wildcard string to a regular expression and then loading
>>> >> >   it into the latter. This allows the nice, fast, wxString-compatible,
>>> >> >   and well tested regex search to be reused for wildcard search.
>>> >> >
>>> >> > --
>>> >> > Chris
>>> >> >
>>> >> > _______________________________________________
>>> >> > Mailing list: https://launchpad.net/~kicad-developers
>>> >> > Post to     : kicad-developers@xxxxxxxxxxxxxxxxxxx
>>> >> > Unsubscribe : https://launchpad.net/~kicad-developers
>>> >> > More help   : https://help.launchpad.net/ListHelp
>>> >> >

Attachment: regexp-and-wildcard-match.patch.gz
Description: GNU Zip compressed data