Saturday, August 23, 2008

Creative Error Messages

My brain is frying. While trying to figure out a bug in Taskerbot -- which mysteriously has assigned just SOME tasks of SOME viewers SOMEtimes to the mysterious userid 9 which does not even exist -- I finally gave up and decided to just remake the whole form/script and do it differently.

I originally coded it in the context of TKR which is such a pain in the butt because of the identity/privacy issues and not being able to just use the USERID all over the place like normal. tBot doesn't have that issue and is going to be standalone so I'm redoing it the way it oughtta be.

While I'm at it I redesigned the top border, and the footer, and so on.

Then I decided my error messages need to display more prettily.

Then I decided that if I'm going to be writing code I could at least be creative.

So my new rule is that all error messages have to be in some form of poetry. By default, haiku, but it could be something else.

This is my sense of humor. Users will have to live with it.

This is the code that calls it (inside a CFIF tag):

<cfset errtype="notice">
<cfset errdetail = "localtask_detail p=step3 IF check viewer tasker must be session identity">
<cfset errmsg = "Avast, Imposter!<br /> You're not tasker, nor viewer!<br /> We hoard it from you.">
<cfinclude template="localtask_error.cfm">

And this is the localtask_error.cfm template:

<cfparam name="errtype" type="string" default="notice">
<cfparam name="errdetail" type="string" default="no code defined">
<cfparam name="errmsg" type="string" default="Programmer failed you<br />No error detail is here<br />Sorry. Try again.">

<div align="center"><table width="600" style="border:0px solid black; color:666666; padding:0px; margin:5px; margin-top:50px; ">
<tr valign="middle"><td align="left">

<fieldset style="clear: both; font-size: 125%; border-color: 990000; background-color: ivory; border-width: 1px 2px 3px 2px; border-style: solid solid solid solid; padding: 30 15 20 20; margin: 4 4 4 4; font-family:Calibri,Candara,Trebuchet,sans-serif; color:black;">

<legend style="font-size: 85%; font-weight: normal; color: 990000; margin: 0 0 0 0; padding: 0 5px;"><span style="color:gray;">tBot TASKER</span>  
<cfif errtype is 'notice'><span style="color:990000; font-size:150%;">Notice</span> <span style="color:silver;">| Error | Omission</span>
<cfelseif errtype is 'omit'><span style="color:silver;">Notice | Error |</span> <span style=" font-size:150%; color:990000;">Omission</span>
<cfelseif errtype is 'error'><span style="color:silver;">Notice |</span> <span style=" font-size:150%; color:990000;">Error</span> <span style="color:silver;">| Omission</span>



<br />
<br/ >
~ dojobots
<br />
<br />
<br />
<small><span style="color:gray; font-size:80%;">#errdetail#</span></small>



While I was at it, I also replaced the main application file error template. This is the one that comes up for CFML or SQL errors (or just before the webserver reboots, or just after it has and the database isn't fully up yet again, or...) to be something a lot more cheerful than the last error page, for both TKR and TBOT:


Thursday, August 21, 2008

CF is Less Girly

OK so the girls can sue me, but I think this image a couple folks use as sig blocks on a forum I frequent is just HILARIOUS. Go Arnold!

Monday, August 18, 2008

Drag & Drop Sortable Lists

I'm working on the ROTA module of Taskerbot now, which I've started and stopped at least three times. Par for the course, I've mostly forgotten whatever the hell I was doing last time I was working on it, so I'm starting over. Again. On the bright side, three years of occasionally working on the idea finally seems to have gelled in the back of my little tiny brain, and now I actually think I can do it, and make it fairly simple in fact, unlike the complex coding behemoth it began.

Tonight the challenge was the basic creation of the sort. The manager takes a list of his members and using simple drag & drop javascript, puts them in the sequence that he wants for event-assignment (usually, tasking).

First of all, there are some precompiled javascripts that help do all kinds of things more easily and more cross-browser compatible, so I went and got those. They are Prototype and Scriptaculous. I got the latest .js (javascript) file from each of those and put it in my directory, and added a head call to each.

<script language="JavaScript" src="../tbotscripts/prototype.js"></script>
<script language="JavaScript" src="../tbotscripts/scriptaculous/scriptaculous.js"></script>

OK, so that is set. (Note that scriptaculous is actually a small library of files; you need to put them all in the same directory, then just call the one.)

Then after searching I found a script (and I can't recall the source or I'd link to it) for the "sortable" feature.

However I could not figure out how to make it work with DIVs as I allegedly can use. Nor could I figure out how to make it write to a hidden HTML form field. After screwing with both of them for quite some time, I finally thought, why am I making this so hard? Yes it would be nice if I had the neat canned javascript for that but I don't, and there is definitely a manual way around it. These items are going to be looped out from a cfquery after all so it's not rocket science.

So I formatted the LI list elements to look like DIVs. And then I put a hidden field inside each of the list elements that would carry the userid (UID) value. When I call the name of that form field, it will automatically give me a comma delimited list of all the values set with that field name, in the sequence they are shown on the HTML form. I tested it and it works just fine.

So I added this to my HEAD section:

li { list-style-type: none; border: 1px black solid; padding:3px; margin:4px;
margin-top:6px; background-color: FFFFCC; color: black; width: 300px; }

And then I added this to my HEAD section:

<script language="JavaScript">
function getOrder() {
var orderList = '';
orderedNodes = document.getElementById("sortable_list").getElementsByTagName("li");
for (var i=0;i < orderedNodes.length;i++) {
orderList += orderedNodes[i].getAttribute('recordid') + ', ';

And here is a simple static version of the code for the sortable list in the page. Note the little script in there -- it must come AFTER the sortable list.

<form name="myform" method="post" action="working1.cfm">
<ul class="sortlist" id="sortable_list" style="cursor: move">
<li id="1" class="sortlist"><input type="hidden" name="uid" value="1">name 1</li>
<li id="2" class="sortlist"><input type="hidden" name="uid" value="2">name 2</li>
<li id="3" class="sortlist"><input type="hidden" name="uid" value="3">name 3</li>
<li id="4" class="sortlist"><input type="hidden" name="uid" value="4">name 4</li>
<li id="5" class="sortlist"><input type="hidden" name="uid" value="5">name 5</li>
<script language="JavaScript">
<input value="Submit" type="submit" name="sub">

Of course if it were dynamic, the output would be more like:

<cfoutput query="myqueryname">
<li id="#myqueryname.uid#" class="sortlist"><input type="hidden" name="uid"

(No wrap in the LI line above--the PRE code puts it outside the display is all.)

>> Here is a full simple demo <<

Anybody should be able to copy the view-source from that and make it work.

Of course, this is actually the EASY part. It's working out all the dates and the looping that is considerably more work. That comes next!


Wednesday, August 6, 2008

Little Import Utility

Me and a friend were talking about setting up Remote Viewing tasks as an experiment, using a number, words, etc. to see if we would get symbol, or be able to dowse for numbers / letters, etc. It turns out the utility I wrote for that last night will server as a lovely improvement to the existing "bulk self-tasking" upload option in Taskerbot.

Basically, in the new utility, you have a 'settings' page where you can write your own 'psychic intent' as a default, and set the visual style you want your feedback in. (So you can make big red letters on yellow background in Impact font, for more effect.) Then you put whatever you want to 'view' -- numbers, words, phrases, whole specific tasker intents, reference to other existing task numbers, whatever! -- in a single column in a spreadsheet (simple) and click to import. It will create up to 250 new tasks on the fly. The system would let you have 3000 chars but spreadsheets will limit it to 250.

The existing 'bulk self-tasking' is a simple form that lets you type in "task number references". But (a) it's all online and (b) it only allows 25 characters. Because that is tying into the 'my task reference# field', whereas the new utility is actually using the text feedback field. And so in this one you could even use html or an img ref or brief EMBED code if you wanted so it could be pretty robust.


The "style" thing is a quick-and-dirty HTML that I made with as much find/replace in Textpad (my beloved Textpad...) as possible:


The combo CFLAYOUT-CFGRID I've been working on for the RV Galleries has not only fried my brain but is at the point of making me cry. The same queries that work just fine on the live site cause a timeout when I try to run them for the grid on my local... I have no idea why. Every time I get something working, nearly every thing I do after that breaks it (and breaks each thing I accomplished before them) and it needs re-re-re-fixing. Some of it's because the middleware just needs more/better development... there are 'quirks' especially in multiple things working together and often stuff just flat out doesn't work, or works fine alone but even though they are supposed to be wholly separate pages, when combined into the tab layout, changing a stylesheet on one screws up another for example. It's never ending. I have to take a brief break from that before I burn out entirely, and do something else that's constructive. So last night it was "Strings" (Numbers-Words-Phrases or NWP) bulk import utility.


Back to the utility: So basically it is using the groovy CF to Excel custom tag that my server company makes available. It's ridiculously easy. Basically you spec the tag, and it saves the content of the spreadsheet as a query set. Then you do a "query of query" for the data. And in my case, then loop it out so you can put a variety of unique/other info into the insert.

<cfx_excel2query r_qResults="importthis" file="#form.importxls#" startRow="1" maxRows="250">

That creates the query set of all the values in the spreadsheet.

<cfquery name="importit" dbtype="query">>
select COLUMN1 as myfeedbackraw
from importthis

This goes and gets just the first column of values, in this case.

Then I loop through. Weird thing, I have scripts used in a dozen forms for generating random task numbers and so on, and they cracked in the loop, giving me all kinds of bizarre results. I couldn't figure out why, but decided not to deal with it, and instead just used a couple generated randoms for both the random task number and the unique 'task reference' number.

<cfset thenow = "#TimeFormat(Now(),'LL')#">
<cfset therand = "#RandRange(32,999)#">
<cfset therand2 = "#RandRange(32,999)#">
<cfset vwrtn = '#therand2#-' & '#variables.thenow#' & '-#variables.therand#'>

<cfset theTN1 = "#RandRange(2001,9999)#">
<cfset theTN2 = "#RandRange(3111,8999)#">
<cfset tasknum = '#theTN1# - #theTN2#'>

That used two randoms and "milliseconds" of a time for insert, to create a 3-3-3 digit 'task reference number' which really means nothing to the viewer but the system needs something unique there. It creates a 4x4 task number for actual assignment (the system task number).

Meanwhile, I use CFDUMP all over the place for quick outputs when nothing needs to be clicked on in that output. But the one on the import page has like a 5 point font. Go figure. I haven't figured out why it's so tiny yet.

Monday, August 4, 2008

Cleaning up the Display

CFGRID, combined with CFLAYOUT (tabbed variant) and CFMENU (simple, not dropdown, as the dropdown does not play well with the CFGRID where they overlap), certainly does make the pages look more clean and professional. And the amount of code in the actual CFM file is drastically reduced, over what I had to hand code (I do everything by hand, old school!) for a paginated list previously.

Old version of the Galleries comments page:

Coming-soon version of that, and the other pages (note the tabs near top):

All that white space at the top right makes me think I need to develop a logo.


My current pages take a long time to load, because they are going back to the database a zillion times for detail. My database server and webserver are both suffering far too heavy a load for the amount of data that's actually happening in the site. There's a lot of stuff that I "should have" done more of up front on that website, like views, stored procedures, and much better SQL that would put more of the processing in the database and less in the middleware, and much better middleware coding, that would minimize the amount of work and time.

The new grids load all the data in at once. The page/tab loads much faster than my ordinary pages, although then waiting for the grid to populate seems like it takes a bit but that's mostly just because instead of waiting for the page to change/load, that happened right away and now you're just waiting on the data. But, once it loads, it's all there. You can do a radical subset or opposite sort and pagination and it is nearly instant -- no waiting another long time for another page to load. So I can not only show the most recent comments in the galleries, but I can let someone click a filter to show comments to them, or from them, or anything from the oldest rather than newest date, and bam there it is. Stuff that would have taken a page load and another wait the old way.

I think not only is this going to end up saving a whole lot of load on my server, but I think it's also going to result in more pages visited, in part due to the speed and in part because I think the tabbed pages make it more intuitive to click through each one, than to 'go find' other pages in the former layout.

The one thing I expect will get to people is that we are all used to putting as many records on a page as possible so we don't have to change pages so often. The grid refreshes only the grid, not the page. That means if you add records so you have to scroll, when you press the 'next' button, the grid shifts--but it doesn't take you back up to the top of the page. You'd have to manually scroll up for that. But the grid is so fast to change pages that it really doesn't matter if there are fewer records in the main display. (I made an option for people to change it.) But I think that will take users some getting adapted to.

In the meantime though, I'll be much more willing to 'claim' the site to other programmers if it looks better. I once made this lovely, clean, elegant, white with blue and silver design there and showed it to a dozen people and like 1 or 2 liked it. All the rest said they LIKED what I called the 'circus nav', different colored buttons along the top of the page, etc. It was just painful. I'd worked out these lovely flash-based forms and everything... ah well. I didn't want to alienate them by making it actually seem SO much more classy and professional that they weren't comfortable in their hokey walmart website anymore, haha! I used to do a site for a guy who ran this multimill company selling satellite TV stuff. He deliberately wanted it to be slightly hokey. He felt that was his audience and their print catalogs had that in mind too. Kinda interesting.

It took me nearly a week of nights and all weekend just to get two pages done seen above, and the filters aren't even working right yet. It's exhausting. But hopefully I've learned enough now -- I'm having to learn a LOT of stuff! -- that the rest of the galleries will go much more smoothly. Ironically I didn't even go in there to do anything in this part of the website. I went in to work on this project called Taskerbot, but got distracted.


Hacking SQL to Save CFGRID: LEN, CHARINDEX, Filename parsing

OK, so then I ran into the same problem I had that I blogged about previously, which is this: getting a full link, image, with unique record ID, into the single column value from SQL so I can have it in my CFGRID.

With one hitch: this time, I needed to use logic to evaluate the image filename. The filename that is in the database is the one for the actual photo, but the filename I needed is a mini-thumbnail that is based on that photo. Now, the thumbnail naming is consistent, at least in regards to matching the database's photo filename. But there are two big caveats, that kept causing SQL errors:

1 - some columns didn't have any value at all there. Their original upload was a PDF for example so there are no files and no value in the database.

2 - the filename entries in that column varied in length. So I couldn't just say, give me the first 5 characters only, then add a new extension (as all the mini thumbnails are JPG, but the filetypes in the database may vary).

This means I had to do four things:

1 - Get the real length of the filename entry

2 - Get what the length would be MINUS 3 digits (the filename extension)

3 - Get only that quantity (#2 above) of characters from the original filename

4 - Then add strings to it, to add the 'mini_' and 'jpg'. And while we're at it, add the strings that will give it the IMG REF and A HREF because CFGRID is not going to be doing that for me either.

5 - And during all this, we have to make sure we are NOT trying to do database math on column values that are incapable of having it done to them or it will probably trigger an error for the user. So if there is no filename or it's too short we just have to do something else.

The first thing I had to figure out was how to stop the logic if the value wasn't up to it. So my first "condition" in my CASE WHEN statement related to that:

CASE WHEN LEN(dsfbfile) > 5

If the filename isn't 5 digits long this is all gonna fail and/or it isn't a real file anyway, so we've taken care of that.

Then I tried to combine the LEN (char length in SQL Server) function and the math subtraction function, to put it only one statement, both 'get the length' and 'remove the length of the extension'. That worked fine.

Unfortunately when I tried to use that it failed, because this is not like CFML, that column name you create is not creating a 'variable' you can just go using everywhere else from then on. It would give me the error INVALID COLUMN every time I tried to use it "as if" it were a number. This is what I tried to do, which failed:

SELECT myrecordID,

LEN(dsfbfile) -4 as THENEWLENGTH,


'mini_' + filetrim + 'jpg' AS minipic

You see in line 3 I was trying to make that new 'thenewlength' column that I had created, stand in place for an actual number, which SQL Server was not happy about. I actually don't think it would have been happy about me using the calculated column in the last line either but I didn't get that far before I changed it.

Then I found reference to this cool function called PARSENAME:

PARSENAME retrieves parts of string delimited by dots.
It is used to split DataBaseServer, DataBaseName, ObjectOwner and ObjectName
but you can use it to split IP addresses, names etc.

DECLARE @ParseString VARCHAR(100)
SELECT @ParseString = 'DataBaseServer.DataBaseName.ObjectOwner.ObjectName'

got that from this address

I wasn't entirely sure how it worked but I got to something else before then so I didn't use it, but thought I'd mention it as it'd be a cool thing I bet for dealing with "filename" manipulation (getting full name or extension easily).

Then I happened to see another example on that same page I linked above, that did something neat I hadn't seen before. First, it used a function called CHARINDEX. Here's a neat article on CHARINDEX and PATINDEX (pattern-index). Basically this function works for 'searching strings' in a database. You can look for a pattern (PATINDEX does wilcards) or something specific like a single character or string (CHARINDEX does that).

Well I didn't know about the use of it in search (I'm still using LIKE % which is probably why my search forms suck), but I saw this example of it used LIKE A NUMBER in a LEFT function -- exactly what I'd been trying to do, but SQL-Server was balking at my using my 'generated value' (or even the logic for the generated value) "as" a number.

This logic, and I'm not even sure why but it does work, looks at my column value, goes to the period in it, goes back one character (-1, to BEFORE the period), and then gets me the string of text that is left of that location, all from the 'imagefile' column exampled below:


I was able to drop that into a string manipulation, and from an input of an 18 character filename (WHEN it exists), output a link, and image (a different image), with the unique record ID on the link, and if the column wasn't a real picture or any value it would write in a filename value, so I can put a 'placeholder' image file there to keep displays from being funky.

So the good news is, here's the code that worked:

SELECT recordid,
CASE WHEN LEN(imagefile) > 5
THEN '<a href="rvg_showpractice.cfm?id=' + CAST(recordid AS varchar(12)) + '"><img ref="' + 'mini_' + LEFT(imagefile,CHARINDEX('.',imagefile)-1) + '.jpg" /></a>'
ELSE 'nofilehere.gif' END
AS 'See Session',
anothercolumn here
FROM table

Now the bad news: it took >59000ms seconds to run on 1000 records. I kept taking the quantity of records down, down, down, until I only had 150 records, and still it was taking like >41000ms.


So I ditched that whole phrase about the mini-thumbnail with link. It is still a long query, even on fairly few records, but it's not so long it's at risk of timeout at least. I will have to put the images somewhere else I guess, where the query is not on a view table made of two other view tables, but something more straightforward so assumed fast.

Still. I was damn proud of getting it to work at all.

Hacking SQL to Save CFGRID: CASE functions

OK, now that I'm using CFGRID that means I can't use all the simple logic that CFML is so famous for making easy.

The irony of this is that it's a big deal how 'easy and lovely' CFGRID is. But it's mostly so freakin easy because it forces you do all the hard work you used to do writing CFML code, instead writing SQL code.

It's rather like a thing at work recently. We used to take content from editorial and build a product. Then we got this automated script tool that builds all the XML almost instantly. Unfortunately it requires the editorial people spend eons making this absolutely perfect, must be precisely formatted, multipage excel workbook which in some cases amounts to 55 sheets of hundreds of lines each... screw up a field and it's a problem, screw up something important and it won't run at all. These people are far less technical than we are, so it puts a larger burden on them. Now production is happy because we can say, "Hey, now we only charge a fraction what we used to, and it's much faster, to build that product." But editorial is doing at least as much work as our PM's used to, except on the front-end, rather than the back end, and frankly they didn't have enough time to begin with (probably why their turnover % is ridiculous). In the bigger picture, the equation and the answer did not improve.

Where was I.

Oh yeah. In my primary coding project there is a gallery of sorts, and the artists (we call "viewers") have the choice as to whether they use their name, their alias, or 'anon' on their contributions (which we call "sessions"). The problem is, this means if you are making a list of the last entered sessions, you don't know what you can put in the space that has the artist's name. Not only that, you don't even know which column it should come from, until you look at the value of a totally different column called REVEAL, that is for their settings.

So you look at REVEAL and say, if it's zero, anon, it's one, alias, if it's two, get both first and last name and combine them. That's all very well in ColdFusion. You have to loop the query out anyway in the page, and you do a CFIF statement with it, so what displays just depends.

<cfif query.reveal is 0>'anon'<cfelseif query.reveal is 1>#trim(query.alias)#<cfelseif query.reveal is 2>#trim(query.fname)# #trim(query.lname)#<cfelse>?</cfif>

But now to CFGRID, ah, yes... it only displays what the database spits out, which means if you want something like a combined field value, a variety of formatting values (with some exceptions), or god-forbid a CONDITIONAL value, you're doomed. The only way to get it is to go back to SQL which feeds the grid and figure out how to get your final answer out of that to begin with.

So this evening I got to learn about CASE functions.
This is SQL Server's version of IF THEN ELSE in most code languages.

It turns out, it's not that hard, once you get the syntax down. At first I thought I could only do two options (like some spreadsheet formulas) but I'm not sure there is even a limit. In the stuff I found online, they never used in their example real data, like the value of a column in that table -- they always used some string, like CASE WHEN saleprice > 500 THEN 'bigsale' or whatever. I wasn't sure using a column value rather than string would work, but it does.

Also, while I'm at it, I go ahead and rename all my columns to whatever I want the title to be in the grid, since then I don't even have to put that into the cfml file, as CFGRID will automatically use the column names if the header isn't defined. And some of my column names are pretty damn obscure (because they may have begun life for a different usage) so this kind of makes the queries clearer in retrospect, since I have the memory of a gnat.

If you want to use a column name that has spaces, dashes etc. just put the name in single quotes.

Here is the logic I used to get the identity column working right:

CASE WHEN reveal = 1 THEN alias
WHEN reveal = 2 THEN fname + lname
ELSE 'anon' END
AS Viewer

So it uses the actual value of columns depending on the value of another, and then uses a string in the last case. And it names the final value "Viewer".

Sunday, August 3, 2008

CFGRID and TEXT WRAP, hover row color and HEADER row style

You would think text wrap would be a parameter on the tag itself. It took me quite some searching to find the answer to that one.

There is actually CSS styling you can do in huge detail on CFGRID, but it isn't provided with the standard documentation, so it is totally metaphysically obscure. None of it seems able to remove the invasive underline should you be crazy enough to use HREF in your CFGRID, but aside from that, everything else seems do-able.

Here are the three little things I critically needed to find and use:


It would just truncate my text in every row. Since it won't let you size the grid by % but insists on a hard coded pixel number, that just made it worse.

Your CFGRID must be in a CFFORM. Give that an ID value. Mine:

<cfform name="myform" id="myformid">
<CFGRID . . .

Then in your stylesheet do the following:

.x-grid-row td {white-space:normal;}

That will make it wrap. The above means it affects everything 'inside' the 'myformid' ID value, and it affects the entire grid, and all the row, and the table cells within the grid rows, and the white-space:normal makes that wrap.


I wanted the header row of labels to look different than the default. I know everyone loves those silver gradient skins but the top row seemed to sort of fade into unnoticed in this case. There is another style that lets you mess with that:

.x-grid-hd-text {background-color: #D9CCFF; color: #661A99; font-size: 120%; font-weight: bold;}

That'll do it.


I'd set all these style values but the entire row color change when you mouse over it wasn't changing, and it was some kind of cyan-like color by default. Looked really interesting with my purple-based grid as you might imagine. I finally found the code to change that row-sized hover color:

.x-grid-row-over td { background:#661A99;}

So there you have it. To set my row styling, text wrapping, and a decent hover, I had this:

<style type="text/css">
.x-grid-row td {white-space:normal; text-decoration: none;}
.x-grid-hd-text {background-color: #D9CCFF; color: #661A99; font-size: 120%; font-weight: bold;}
.x-grid-row-over td { background:#661A99;}

And it looks like this:

Hacking SQL to Save CFGRID: Image as column value, HREF link with RECORD ID

This is only part of solving a larger problem. The problem is using the HREF on CFGRID underlines every damn thing and I couldn't figure out how to get rid of it. I did find extensive notes on CSS styling for the grid, down to every single cell, but nobody crying about it on the internet had gotten any answer on how to get rid of the underline.

So I took the HREF parameter off my CFGRID because the underline looked so horrible and made it so unreadable I wouldn't even use the grid if I had to suffer that. Which left me with the following dilemma: since the grid no longer links to something semi-automatically (posts to a form, click chooses a whole row, you can pass the ID value you need), I would have to link to it manually.

As part of doing this I decided that much like the simple HTML table I've been using prior to trying to implement CFGRID, I would use a little icon that would intuitively tell people (a) to click that for a link, and (b) that it would open in a new window. So I had three things to figure out:

1. How to get an image as a column value, because CFGRID only lets you insert a query column name, not html, and

2. How to get an html HREF link into a column value, for the same reason, and

3. How to get the unique record ID into that link.

To get a single (just one, same image on every record) into a column value, I did the below. Note that if you need a different image for each record, and you can build that filename/loc from whatever values are in the query, you could do that too. Use the logic of the Record ID link below, and modify that.

SQL Server uses + to concatenate strings and columns. Other DBs may use things like CONCAT function. You use single quotes around every string, and no quotes around actual column names.

You have to render, aka CAST or CONVERT, integer fields to strings before you can concatenate them.

'<a target="_blank" href="myactionfile.cfm?id=' + CAST(p.myid AS varchar(12)) + '">' + '<img src="../shared/icons/external2.gif" border="0" /></a>' as 'See',

then in CFGRID, I just did

<cfgridcolumn name="See" width="35" display="yes" >

and that's it! That changed my ID to a string, and it tied it into a link and an image, so now in my column it looks like this:

You click on the little image and it opens that record in a new page.

Hacking SQL to Save CFGRID: Date Format

In the HTML format CFGRID, my date "mask" formatting, using the CF documentation, didn't seem to work at all. I don't know why but my impression from cfdocs was that it wasn't going to work in the HTML format grid I was using. I hacked this by changing my SQL to begin with, rather than doing it in CF.

Formerly it was:

SELECT mydatefield as DateEntered


mask="EEE DD-MMM-YY H:NN A">

Which sucked and continued to show me whatever it wanted, rather than the "mask" attributes the cfdocs had. So I changed it to this:

SELECT CONVERT(VARCHAR(11), mydatefield, 106) AS DateEntered



And the resulting output looks like this:

22 Jun 2008

You can google for SQL DATE FORMAT or something like that and see the whole huge list of numbers (mine 106, above) that will format dates differently.

CFGRID and CFC Binding How To

I avoided using a CFC binding initially because I didn't grok how and, before I knew that my route made pagination impossible, a query in the HTML CFGRID seemed so much easier. But here is how the binding part is done.

In the CFC file you're going to have the following:

1. A function name with arguments
2. Your query
3. Not always but in my case, a query return for the grid.

The function is going to have remote access and return a structure. Here is my code for it:

<cffunction name="comments" access="remote" returntype="struct">
<cfargument name="page" required="true">
<cfargument name="pageSize" required="true">
<cfargument name="gridsortcolumn" required="true">
<cfargument name="gridsortdirection" required="true">

Page means, "Do you want this to do paging, where things are broken into separate pages?" PageSize means, "How big (how many records) should each page be?" The sorts are fairly obvious I think...

If you have the above it means you will have to have the page and pageSize defined in your CFGRID tag in your CFM page, and you can control it from there. I did not define my grid sorts and it didn't break.

Next you have your query. Whatever query you were using to output records prior to CFGRID will probably work fine, but I ended up having to hack mine to do all kinds of stuff that CFGRID made a royal pain in the butt, so I fixed it through SQL instead of CFML. Will cover that in another post.

Then you have your query "return". You find this in documentation under functions, queryconvertforgrid function.

In my case it has three values:
1 - the name of my query
2 - page
3 - pagesize

And I mean those strings (page and pagesize) not true/numeric. Here's my code, where my query is called selectAll:

<cfreturn queryconvertforgrid(selectAll,page,pagesize)/>

Now to put the CFC together you'll want these three things in order. I have a file called getcomments.cfc and here is the code, with the query revised for public use of course. Note that the CFC did not find my datasource which is a variable set in my Application.cfm file, so I had to manually CFSET that value before running the query.

<cffunction name="comments" access="remote" returntype="struct">
<cfargument name="page" required="true">
<cfargument name="pageSize" required="true">
<cfargument name="gridsortcolumn" required="true">
<cfargument name="gridsortdirection" required="true">

<cfquery datasource="myds" name="selectAll" maxrows="1000">
My query goes here

<cfreturn queryconvertforgrid(selectAll,page,pagesize)/>

Then in my CFM file, my HTML format CFGRID used this code to call it:

<cfform name="myform" id="myformid">
name = "myGrid"
<cfgridcolumn name="id" display="no" >
<cfgridcolumn name="DateEntered" display="yes" header="Date" mask="EEE DD-MMM-YY H:NN A">
<cfgridcolumn name="Alias" italic="yes" display="yes" header="From">
<cfgridcolumn name="See" width="35" display="yes" >
<cfgridcolumn name="theComment" fontsize="1.2em" textColor="##661A99" width="700" display="yes" header="Comment">

Notice that there is some styling in one of the columns and that works, and date formatting in the date column which did not work at all so I did something different, which I'll cover in another post. The 'see' value there is a big string from a special SQL call that gave me an HREF link, and an image, and appended the record ID, because stupid CFGRID wouldn't let me get rid of underlining every damn thing if I used the HREF attribute.

CGRID and Auto Pagination

Everyone raved about how cool it was that ColdFusion's CFGRID tag automatically would do the pagination for you.

All the little examples I looked at, had that nifty little page number and button at the bottom.

Except mine. I could do code pretty much just like the examples, except that I was using an HTML form and a QUERY in it, but the little pagination thing never appeared.

It turns out that you cannot do pagination in a QUERY HTML CFGRID in CF8. Instead, you have to put your query out in a CFC page and "bind" that to the CFGRID. It is the CFC page that actually works out the pagination stuff.

I'll cover the CFC and binding separately.

The Blog

Yes. I know this is a canned design theme and I don't care to spend the time making one pretty like my other five blogs on other topics.

Since I only do coding on my "extra" hours -- read: hours I should be sleeping, not working for free for hobby sites -- my time to figure things out is limited. And since I only use Cold Fusion (CFML), HTML and CSS -- I do not know jack about javascript except what I can steal, and I don't know much about any other kind of coding, nor have I time to learn -- this state of semi-illiteracy in programming means that every smallest problem is a true Day In Hell for me. I don't know how to script hack workarounds. And ColdFusion's documentation, while quite robust in places, in others leaves something seriously to be desired.

I have googled more than any human should have to, to find solutions to my Cold Fusion coding problems. What I often find is

a) A solution by accident, e.g. someone is doing something entirely different but after enough of those I chance on an idea that I try and it actually works for my own problem too, and

b) An internet of people bitching about some problem but absolutely nobody who has posted an answer or solution to it.

I finally decided that although I will only post occasionally, I'm going to post my issues and solutions or workarounds on a blog, so the next poor sucker trying to google for answers on these problems might actually find MY blog and get one.