#foswiki 2014-01-21,Tue

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)

WhoWhatWhen
***ChanServ sets mode: +o pharvey [00:16]
..................................................................... (idle for 5h41mn)
ChanServ sets mode: +o pharvey [05:57]
Rich_MorinThe pull-down menu in the "Create New Task / Bug Item" popup has several options: .*, Engine, Extension, and Web Site. The first option isn't defined in the menu, but Marketing (which is unavailable) is. Wassup? [06:01]
......... (idle for 41mn)
***ChanServ sets mode: +o CDot [06:42]
..... (idle for 24mn)
ChanServ sets mode: +o pharvey [07:06]
.... (idle for 19mn)
Rich_MorinSvenDowideit: Yet another unsolicited suggestion - http://foswiki.org/Tasks/Item12725 [07:25]
***ChanServ sets mode: +o pharvey [07:34]
................... (idle for 1h34mn)
Colasping
ooops wrong chan
[09:08]
CDotRich_Morin: it's a delicate balancing act. We *could* create an "everything and the kitchen sink" distribution, but maintenance would be a nightmare - it's difficult enough already with the limited number of plugins we do ship by default. We try to strike a balance where the "most useful" plugins are included by default, and installing others is as easy as possible (ok, so we don't always get this right). But we just don't have the (human) resources [09:23]
............. (idle for 1h2mn)
MichaelDaumit isnt a human resource problem only, though any other aspect becomes academic when this prerequisite fails.
it is like hybrid cars: they are mediocre at best in both leagues.
[10:25]
.... (idle for 16mn)
GuilainCGuilainC precise to MichaelDaum_, that my work is to make, with the help of my team and foswiki ;)... an hybrid car... ;) [10:42]
MichaelDaum:) [10:43]
.... (idle for 16mn)
***ChanServ sets mode: +o Lynnwood [10:59]
...... (idle for 25mn)
SvenDowideitRich_Morin as i replied to your task
its trivial for you to do - start with ``apt-cache search foswiki | sed 's/ - .*//' | xargs apt-get install``
as you identify the need, you should also be able to find the needy who would use it - and thus the testing CDot mentions would be done
if you can't identify the needy, then perhaps you're talking of a solution in search of a need?
SvenDowideit has done an awful lot of talking up a 'need', only to realise over time that if no-one is clambering to help, then the 'need' is imagined/wishful, rather than real
[11:24]
........... (idle for 50mn)
GuilainCi making dataform with the use of AutoCompletePlugin, and i want to get wikiname written with global link like "%USERSWEB%.JohnDoe" and not "JohnDoe". I've made some try, but without succes. How to do it ?
datatopic="Sandbox.AutoCompletePluginTestData" datasection="usernames" delimchar="," itemformat="'Main'+item[0] + ' (' + item[1] + ')'" is no working
[12:20]
........... (idle for 54mn)
***ChanServ sets mode: +o gac410 [13:15]
.... (idle for 15mn)
gregg4567 has left [13:30]
.... (idle for 18mn)
tsnfooSomeone was asking about a vector-based logo the other day. I uploaded an SVG version here: Community.BrandLogoTalk
http://foswiki.org/Community/BrandLogoTalk
[13:48]
CDottsnfoo: nice! [13:53]
JulianLevensDitto [13:53]
tsnfooCDot: thx. Thought it would help :-) [13:53]
CDotI always liked that logo; but persuading certain die-hards that it's the best was a real challenge :-( [13:54]
tsnfooIt looks very modern, very recognizable. [13:55]
.... (idle for 16mn)
MichaelDaumI've uploaded a foswiki logo with an improved font. [14:11]
Lynnwoodlooks very nice [14:11]
MichaelDaumthis is good old arial instead of the serif font.
the other derived images need adjustments as well (favicon, configure)
[14:11]
tsnfooHmm. The sans font looks a little shout-isg to me, agh hat weight and proportion.
shout-ish, rather.
[14:27]
MichaelDaumshout-ish? [14:30]
.............. (idle for 1h7mn)
CDotFWIW I like the crispness of the sans, but it's a rather hackneyed meme. [15:37]
....... (idle for 31mn)
MichaelDaumthe serif font is difficult to integrate into a web design or other artwork.
and become fuzzy on lower res (favicon, badge)
[16:08]
JulianLevensWhy do we not use the new icon as a Favicon?
That would also decouple the matching text allowing flexibility regarding the matching FOSWIKI text
[16:10]
gac410Is it necessary to choose a font, vs. using something like fontastic, or a jquery plugin, to let the user select whatever they want?
With SVG, can't the logo font be consistent with the rest of the site?
gac410 asks something that he doesn't really know anything about.
[16:12]
CDotbefore I google, is there any way to do a TEXT to INT dynamic type conversion during a search?
sorry, should clarify; I mean in SQL
[16:14]
JulianLevensCDot: There are usually convert functions, but I suspect they will be non-standard [16:15]
CDotnon-standard is OK, I can use the personality module. OK, thanks, I'll google. [16:15]
JulianLevensCDot: I thought so, MySQL it's automatic [16:19]
CDotAutomatic? Ouch = so "1" < "1.1" will automatically cast to INT? [16:21]
JulianLevensCheck the docs carefully and test. There must be some limits to automatic [16:22]
CDotautomatic is what Foswiki expects, but as you say there will be limnits.
hmph. Looks like I need to add a CAST operator to the FW query syntax (should have done that a long time ago).
[16:22]
JulianLevensI already proposed something similar for date(). You either need a matching number() operator and/or a generic cast
http://foswiki.org/Development/DateAsAQueryHint
[16:26]
CDotdebatable whether a cast is the right approach. It's not the *operands* that need the case, it's the *operator*
^case^cast
e.g. X < Y is numeric, X lt Y is lexical
X<Y can be SQLed as CAST(X AS int) < CAST(Y AS int)
[16:28]
JulianLevensNot sure that's necessary. If X < 1234, then rhs is clearly numeric [16:30]
CDotwhat if X<Y?
fields[name="FieldOne"].value<fields[name="FieldTwo"].value
[16:30]
JulianLevensOTOH if X < Y and your not sure of the type of anything then you need a hint (i.e. not a cast) of one side or the other
fields[name="FieldOne"].value<number(fields[name="FieldTwo"].value)
It's down to a feel of what works best for users
[16:31]
CDotthat's equivalent to casting the op (just a bit more work)
I think being realistic about it, the only op that is a problem is =
because < and > are pretty much always numeric
[16:32]
JulianLevensBut alas not always [16:33]
CDotbut 0.0=0 needs to compare numerically, and "0.0"="0" lexically [16:33]
JulianLevensHow will this affect exiting code if = assumes numeric [16:33]
CDotthe existing code uses late typing; so it *knows* the type of values it is about to compare [16:35]
JulianLevensY, I just worked that out [16:35]
CDotCDot is kicking himself for doing that [16:36]
JulianLevensIIRC that drove my proposal (albeit originally for dates) [16:36]
CDoty, but I feel your proposal layers too much complexity, compared to the simple expedient of demanding explicit typing [16:37]
JulianLevensNB: It's also fundamentally broken, I came up with a case where a < b < c but a > c [16:37]
CDotyou actually wrote a<b<c? Well, that will surely break
a < b AND b < c and a > c would be an error, sure.
hum. Well, I can't add ops to the query language willy-nilly.
[16:38]
JulianLevensI mean I came up with real field values a, b and c (2.3 2.0% 2.0 or something like that) where 2.3 < 2.0% < 2.0, but 2.3 > 2.0 [16:40]
CDot"2.0%" will be interpreted as a lexical value
double hum. I *can* add ops to the query language from a contrib.
[16:40]
JulianLevensIgnore those values, it was something close to that, it was a deliberate mix of lexical and numbers with no clear order with late typing [16:42]
CDotJulianLevens: how do you feel about my adding gt, gte, lt, and lte to the query language for lexical comparison?
and eq of course
[16:43]
GithubBot[foswiki] FoswikiBot pushed 1 new commit to master: http://git.io/Jo5aoQ
foswiki/master 9c95dc4 MichaelDaum: Item12713: improvements to user profile pages...
[16:44]
***GithubBot has left [16:44]
FoswikiBothttp://foswiki.org/Tasks/Item12713 [ Item12713: new default design ] [16:44]
CDotargh, n.m., eq makes it moot. That *most* be late typed.
must
[16:44]
JulianLevensNo, it doesn't solve the problem for dates [16:46]
CDottriple hum. I think you are right. Unless one of the operands is explicitly cast to an int, we have to assume lexical comp
so your number() op is the right way to go
and date() of course
[16:47]
JulianLevensIn principle it also allows for further types to be added [16:47]
CDotyeeeeessssss.... need to think about that [16:48]
JulianLevensDefault is lexical comp as you suggest, unless X < 1234 type situations or even X < '2003-12-13' although that case requires a string('2003-12-13') just in case it's meant to be a lexical [16:49]
I think from a versatile POV I'd like not to have explicit conversion Duck types et al, just pass the SQL thru when I can, but I'll think about that some more
Think about how the impact of FW gaining strict types in the future. It's probably no more than handling date(number_field) as an error, but ...
[16:55]
CDot"just pass the SQL through" is another matter. I was thinking of doing that - supporting an "sql=" parameter on %SEARCH
or rather, sql_where=
however, big problem with that is SQL dialect, so it would come with a HUGE health warning
duk typing would be lovely, if we could only persuade SQL.....
[17:05]
MichaelDaum%SEARCH{type="sql" ... %SEARCH{type="lucene" ... %SEARCH{type="solr" ... %SEARCH{type="elastic" ... [17:08]
CDotyup. and type="dbcache" ;-) [17:08]
MichaelDaumnow if only the result formating wasn't so iffy
in %SEARCH
[17:09]
CDotthat might provide some focus for cleaning that code up
cos it has a lot of (putting it politely) sub-optimal features
[17:09]
MichaelDaumbesides results sets being a lot more structured than only listing results in solr (facets, highlightning, spell corrections, match groups) [17:10]
CDotyeah, well, let's walk before we try to run
though having said that, on my tests the formatting is now the critical path for searching :-(
[17:11]
MichaelDaum%SEARCH in its current form is quite a needle's eye for more powerful backends.
I can imagine
[17:11]
JulianLevensThrow away SEARCH and start again! [17:12]
gac410gac410 as RM has nightmares about Item12725 [17:12]
MichaelDaum+1 [17:12]
FoswikiBothttp://foswiki.org/Tasks/Item12725 [ Item12725: Needed: "batteries included" Foswiki distribution(s) ] [17:12]
JulianLevensSee Also Item12721 [17:12]
FoswikiBothttp://foswiki.org/Tasks/Item12721 [ Item12721: Support multiple attachments at one time ] [17:13]
CDotSQL search: 10345 results in 0.2 seconds. Formatting: 36.5 seconds. [17:13]
MichaelDaumCDot, you better don't show 10345 lines of html to any user, let alone pulling it thru the net to a poor browser chewing memory [17:13]
CDotnot html; 10345 *results* (i.e. topics) [17:14]
MichaelDaumI was thinking of formatting in terms of rendering results [17:14]
gac410Hi JulianLevens, yeah I've been watching both. The point is a good one, but I think we need a scalpel, not a butchers knife. [17:14]
CDoty, I know. This is a test set, remember. [17:14]
JulianLevensI'd love to create a new SEARCH from scratch, but Versatile was a big enough project, I need a holiday after that! [17:15]
MichaelDaumCDot, rendering 10k topics is bottle necked somwhere, presumably ACLs. [17:15]
gac410A careful inclusion of another round of core extensions would be good. But not *every* extension [17:15]
CDotdunno; I'm focusing on the search ATM. Might look at formatting later. [17:15]
JulianLevensgac410: yeah that's why I deferred to the FoswikiCamp so we can discuss what to include (or not) [17:16]
CDotgac410: can we remove some too? [17:16]
MichaelDaumgac410, extensions is not the right way to think about this all. it should be more of what foswiki should be used for. [17:16]
JulianLevensFoswikiFor... [17:17]
MichaelDaumif it needs no extension, fine. so is 10 extensions. as long as the use case is met. [17:17]
gac410CDot's restful config might make "plugin batch" behaviour better. I can picture a NatWizard that gives you one page that walks through enabling all of the interdependencies for a NatSkin [17:17]
CDoty, that's a thought [17:17]
JulianLevensNote that we could deliver FW without configure (eventually)
But would we
[17:18]
CDotsure. With my work, "configure" is a plugin. An optional plugin. [17:18]
JulianLevensWe could even ship without a default Store, but would we? [17:19]
MichaelDaumUpdatesPlugin needs integration into configure in some way. [17:19]
gac410I can't see it being delivered without some configuration tool, but as topic based wizards, not as one monolithic tool
gac410 has very mixed feelings about UpdatesPlugin
[17:19]
CDotoh bugger, I just remembered I was going to pull attachments into the cache DB using StringifierCOntrib. Is that a good idea or not? [17:19]
MichaelDaumdepends what for [17:20]
JulianLevensWell I plan to cache attachments Strings into Versatile, for local use that's very important [17:20]
CDotCDot has in mind supporting queries over attachment content [17:20]
MichaelDaumstringified attachments, that's potentially a lot of bits to store. [17:21]
gac410If we ship lots of extensions by default, we also need a way to hide the System topics of inactive / undesired extensions to avoid confusing the average user. [17:21]
CDot%SEARCH{"attachments,content=~'Burble' ... [17:21]
MichaelDaumsearch engines normally only index these strings, but don't store it in full. [17:21]
CDotactually, it's pointless. %$SEARCH is fundaentally topic based.
so the best you could do is return a topic that has an attachment that has that content
which would be silly.
[17:21]
MichaelDaumas long as it shows an url to the orig pdf ... [17:22]
CDotBetter to let Solr et al handle that, [17:22]
MichaelDaumsome sql dbs come with fulltext features [17:23]
CDotmsot of them do [17:24]
MichaelDaumgotta run. see you tomorrow. [17:25]
JulianLevensy, I'm off to practice hitting little white balls [17:25]
CDotice swimming, huh? ;-) Have fun [17:33]
................................................... (idle for 4h13mn)
***ChanServ sets mode: +o pharvey [21:46]
......... (idle for 42mn)
farsel'ello [22:28]

↑back Search ←Prev date Next date→ Show only urls(Click on time to select a line by its url)