Jump to content

Module: Matrix Fieldtype & Inputfield


kongondo

Recommended Posts

@kongondo, is there any need to store the empty results?

Just wondered as once you get into the world of combinatorials things quickly get memory hungry big; and PHP associative arrays aren't exactly the world's most efficient storage format. Any idea how this fieldtype works out memory wise? 

Btw, in addition to my previous response, for the curious, at the input level, each individual intersection is HTML named as follows:

<input type="text" name="matrix_products[R5284_C5271]" value="100">

Where R1234 = row page->id and C1234 is column page->id. So, the two dimensional array shown in the post above is for grabbing data from the db, manipulating it as R1234_C1234 intersections and displaying (Inputfield) it to the user in the matrix table cells. On save, we are dealing with a one level associative array as shown in the HTML code above.

Link to comment
Share on other sites

Update: versions 0.0.4 - 0.0.6

Versions 0.0.4 - 0.0.5: Code changes.

Version 0.0.6:  Added ability to copy-paste CSV values to save to current page

Wasn't sure whether ability to import from CSV values in this manner really belonged in a Fieldtype and was wondering whether I should make the feature configurable optionally available? Then again, it makes sense in a fieldtype whose primary purpose is to store excel/grid-like data! Anyway, for now it stays...Will think about adding feature to import from csv/txt file...

Copy paste CSV Data

post-894-0-85844100-1419303855_thumb.png

Edited by kongondo
  • Like 6
Link to comment
Share on other sites

I have done a couple of tests in respect of memory (memory_get_usage) and timings (Debug::timer()) when processing 1000 copy-pasted CSV values in a 100X10 matrix table:

For a fictitious investment company, I generated (using a local install of datagenerator) CSV data consisting of 100 rows (account manager) x 10 columns of customer details containing their:

Full Names, Age, Country, Bank Account Number, Phone Number, Credit Card CVC, Email, Income, Address, Post Code

Btw, this showcases another use for this Fieldtype. It is not only about colours vs sizes  ;)  :P . If you have data such as above that is most likely already available in CSV or similar format and that would not readily lend itself to being converted to pages (even for use in Page Fields), then you probably want to use FieldtypeMatrix...

OK, back to our measurements...I imported my 1000 CSV values into my matrix and got the following results. Note that I am on an i7 windows machine. lightning.pw has been down for me the last couple of days so wasn't able to test there. I am not good at such testing so if anybody else wants to help out I'd appreciate it. I also had to increase my max_input_vars from the default 1000. 

Saving 1000 pipe (|) delimited CSV values

//Note: SEQUENTIAL PROCESSING (I skipped timing methods like get() and set()
//not sure why ___sleepValue() was not showing here! See next test where we exit at ___sleepValue()

11187192: Memory Usage Start: get()Inputfield()
0.0360: Debug Timer End: get()Inputfield()
11270704: Memory Usage End: get()Inputfield()

11772248: Memory Usage Start: ___wakeupValue()

11772456: Memory Usage Start: get()blankValue()
0.0000: Debug Timer End: get()blankValue()
11773896: Memory Usage End: get()blankValue()

0.0700: Debug Timer End: ___wakeupValue()
13372176: Memory Usage End: ___wakeupValue()

12871408: Memory Usage Start: sanitizeValue()
0.0000: Debug Timer End: sanitizeValue()
12871408: Memory Usage End: sanitizeValue()

14262896: Memory Usage Start: ___render()//render calls mergeMatrix()

14263624: Memory Usage Start: mergeMatrix()
0.2190: Debug Timer End: mergeMatrix()
15165696: Memory Usage End: mergeMatrix()

0.2800: Debug Timer End: ___render()
15231376: Memory Usage End: ___render()

Same as above (+sequential) but exiting at ___sleepValue()

//exit before ___sleepValue() returns $values

11707344: Memory Usage Start: get()Inputfield()
0.0020: Debug Timer End: get()Inputfield()
11790872: Memory Usage End: get()Inputfield()

11791416: Memory Usage Start: get()blankValue()
0.0000: Debug Timer End: get()blankValue()
11792856: Memory Usage End: get()blankValue()

11793040: Memory Usage Start: sanitizeValue()
0.0000: Debug Timer End: sanitizeValue()
11793040: Memory Usage End: sanitizeValue()

12762576: Memory Usage Start: ___processInput()

12763296: Memory Usage Start: mergeMatrix()
0.0540: Debug Timer End: mergeMatrix()
13372952: Memory Usage End: mergeMatrix()

13361256: Memory Usage Start: ___processInput() - csv only

13361848: Memory Usage Start: processCSV

13362040: Memory Usage Start: mergeMatrix()
0.0080: Debug Timer End: mergeMatrix()
13486200: Memory Usage End: mergeMatrix()

13613512: Memory Usage Start: get()blankValue()
0.0000: Debug Timer End: get()blankValue()
13614504: Memory Usage End: get()blankValue()

0.0830: Debug Timer End: processCSV()
15346208: Memory Usage End: processCSV()

0.0830: Debug Timer End: ___processInput()
15098552: Memory Usage End: ___processInput()

14992888: Memory Usage Start: sanitizeValue()
0.0000: Debug Timer End: sanitizeValue()
14992888: Memory Usage End: sanitizeValue()

15128552: Memory Usage Start: ___sleepValue()
0.0160: Debug Timer End: ___sleepValue()
15511096: Memory Usage End: ___sleepValue()

Loading the matrix table with the saved 1000 values

//SEQUENTIAL

11763208: Memory Usage Start: ___wakeupValue()
0.0890: Debug Timer End: ___wakeupValue()
13363016: Memory Usage End: ___wakeupValue()

14253776: Memory Usage Start: ___render()

14254504: Memory Usage Start: mergeMatrix()
0.2250: Debug Timer End: mergeMatrix()
15156544: Memory Usage End: mergeMatrix()

0.2680: Debug Timer End: ___render()
15222208: Memory Usage End: ___render()

Don't know what to derive from these results or if I did them correctly :-). From a timing point of view, the results don't look too bad, or?

Saved data screenshot

post-894-0-27798300-1419419473_thumb.png

  • Like 6
Link to comment
Share on other sites

Just for fun, I decided to try saving from copy-pasted CSV data using MySQL LOAD DATA INFILE.  Now this MySQL feature is something you use for really serious transactions. Testing with 'only' 1000 values seems an overkill. Anyway, here's the results:

Testing with same data as in my post above (10X100 table with 1000 values)

Method 1: CSV + MySQL LOAD DATA INFILE

1. Copy-pasted CSV values in the format:
Name|Age|Country|Bank Account|Phone|Credit Card|Email|Salary|Address|Post Code
Quinlan T. Romero|36|Burkina Faso|RS31212461883550021066|(989) 462-3421|863|Integer.vulputate@Curabiturut.ca|£066.39|P.O. Box 161, 2347 Donec Street|6518

2. Flip the CSV values to match our db structure, i.e.: data(matrix_row) | matrix_column | matrix_value
	- explode csv line by line
	- str_getcsv csv line to create array of the csv line
	- push sanitised csv values into a new CSV array mirroring our 'db structure' (i.e. including row and column page->id)
        - we do this because our raw csv data is in matrix table format and LOAD DATA INFILE won't read it correctly
	
		array( 
				[0] => (
				
						[0] => row->id
						[1] => column->id
						[2] => value
				
				
				)				
		)
	
3. Write CSV values to a data.csv file
	- fopen a data.csv in this page's /assets/files. Force create the file w+ if one doesn't exist
	- fputcsv the CSV array line by line
	
4. LOAD DATA INFILE
	- instantiate a new PW (PDO) $this->wire('database');
	- Delete this page's existing values from db
	- Execute a LOAD DATA INFILE, reading through our data.csv and writing to db (at high speed )
	- Delete data.csv
	

Method 2: CSV + 'Saving Normally' 

1. Similar to above but we don't write to a data.csv nor use LOAD DATA INFILE
2. Prepare csv array using str_getcsv similar to above...but
3. Our CSV array is slightly different (but timings don't matter because of the different structure)

		[row->id] => (
		
					[column->id] => value		
		
		)

Results: Memory usage and Timings

                Normal			LOAD DATA INFILE	
				
processCSV()
mem_end		15,105,984		13,767,352	
mem_start	12,754,968		12,761,496	
Diff	 	 2,351,016 		 1,005,856 	
				
time	 	   0.1240 		    0.0740 	
				
___processInput()		
mem_end		14,974,128		13,328,584	
mem_start	12,754,128		12,760,504	
Diff	 	 2,220,000 		   568,080 	
				
time	 	    0.1240 		    0.0940 	
				
				
___sleepValue()		
mem_end		15,504,760		Not Applicable	
mem_start	15,122,112			
Diff	 	   382,648 			
				
time	            0.0160		Not Applicable	

As you can see LOAD DATA INFILE is the (clear?) winner.

Not sure if LOAD DATA INFILE is enabled everywhere though? Anyway, for now, not sure if this is the route I should take. Besides, I got something else planned (in my head!) for handling really large grid tables.. :P

:D Before you ask, no we are NOT using LOAD DATA INFILE with the LOCAL option (security concerns with that option!)

Edited by kongondo
  • Like 3
Link to comment
Share on other sites

Hmm, or maybe I should include this in the present module but with a max 3-way relationship? Otherwise it get messy if more. The 3rd relationship (materials in your case) would only kick in if user configures it in the Fieldtype settings as you suggested. What do you guys think?

I obviously would love to see this being added :) But it might be error prone on the input side of things. You'd have to tell the user exactly what's allowed and how to enter it. I think this needs a bit more thinking on how it can be best integrated.

  • Like 1
Link to comment
Share on other sites

@everfreecreative, you mean besides having a cooler name? :P.

As LostKobrakai pointed out, it's all in the first post. You can't really create a 2D/Matrix table using repeaters, since, er, the rows just keep repeating with the same thing. And if you could create one, I think you would have more overhead using  a repeater to create, say a 10x100 table than this Fieldtype. To understand what the Fieldtype does, think an Excel spreadsheet with column headers and row labels whose intersections contain unique cell values (by unique I mean each table cell can only ever represent a unique combination of two factors different from any other cell). The main purpose of the Fieldtype is for storing and retrieving such values.

As for Table, that's a whole different beast whose scope is far greater than what this Fieldtype does...

  • Like 1
Link to comment
Share on other sites

Did you read the first post here? There's an explanation for exactly your questions.

Yes, I did. The price example confuses me because I'm not sure how it's being used on the front end. Is this being output as a table on the front end for users to be able to determine the price of something they want to order on the site? Or are they using inputs on the front end to select the options they want and the matrix is just used on the back end to determine the price based on those inputs?

I'm used to the idea of a product option increasing or decreasing the price based on some fixed value or percentage, so maybe that's part of my confusion.

It looks very cool though :)

Link to comment
Share on other sites

I obviously would love to see this being added :) But it might be error prone on the input side of things. You'd have to tell the user exactly what's allowed and how to enter it. I think this needs a bit more thinking on how it can be best integrated.

You are right. The more I think about this, the more I see it as a separate/bespoke module...

Link to comment
Share on other sites

As for Table, that's a whole different beast whose scope is far greater than what this Fieldtype does...

While Matrix may not be as flexible as Profields Table, one cool thing (although there are lots of cool things about it), is that your site editors can add new columns to the table - you can't do that with Table fields ;)

  • Like 1
Link to comment
Share on other sites

Also, for some reason in the Fieldtype' Details Tab (see screen below), There seems to be a clash betwen two InputfieldSelects. The selected and saved value of the second InputfieldSelect is not remembered (i.e. no <option selected=selected></option>. However, the value is correctly saved to the database. Some JS clash? Help please, thanks!

I found the problem :)

You are overwriting $field so the value of $field->columnLabelFieldName is blank.

This line is the culprit: https://github.com/kongondo/FieldtypeMatrix/blob/master/FieldtypeMatrix.module#L168

foreach(wire('fields') as $field) {

If you log $field->columnLabelFieldName before that, it returns the selected value as expected.

Hope that helps!

  • Like 1
Link to comment
Share on other sites

Changed this to Beta version. I'd appreciate testing (especially character encoding issues and using CSV/Text tab delimited files generated in other systems other than Windows - Linux, Mac, etc..), thanks!

Changelog

Version 0.0.7
Added optional feature enabling fast import of CSV data using MySQL's LOAD DATA INFILE (off by default, set in Details Tab of Fieldtype).
 
Version 0.0.8
Option to populate matrix table via a .csv/.txt file upload.
Top and Bottom Reset buttons to clear all matrix values before save (handy when you want to restart from a clean slate).
Fixed issue where saved column header label was not being selected in the InputfieldSelect in the Fieldtype's 'Details' Tab - Thx @Adrian
Changed status to Beta.
Version 0.0.9
Fixed a character encoding issue regarding fopen (e.g. £ sign not being displayed) - note: if you really need to save such characters in the db, then you are better off copy-pasting your CSV values rather than uploading (i.e. avoid fopen).
 
==============
 
I'll be writing more advanced find examples for this field using my customer data example. e.g. find records where a customer is <=45 and lives in Cameroon.
  • Like 4
Link to comment
Share on other sites

@kongondo, this looks great, thanks for making it available!

Will take a closer look soon, but just a (very) minor note at this point: I can't help wondering why the subfields are prefixed with matrix_ -- doesn't honestly matter that much, but I fail to see much value in that either, just slight increase in verbosity compared to non-prefixed alternatives  :)

Thanks @teppo. Yeah, I battled with this one; I am not a fan of such prefixes myself :-). I was trying to avoid collision with MySQL reserved words - row(?), value and column. Funny though that in my tests selectors still work when I drop the 'matrix' prefix...hmm. 

Guys..suggestions for naming the row/column/value  COLUMNS? Maybe:

mrow - mcolumn (or mcol)  - mvalue

OR maybe...

use the PW data COLUMN for 'values' (i.e. intersection/combination cell/value of row vs column), i.e.

$page->product.mrow//rows
$page->products.mcol//columns
$page->products.data//their data/values

Any thoughts? Thanks.

@kongondo: had to check, and apparently column is a reserved word, row and value should be fine. Either way, prefixed values make sense after all, no point in having only one prefixed. Also, I kind of prefer matrix_* over m*, and matrix_value over data, so wouldn't chance anything to be honest :)

Revisiting this:

Duh! I forgot that similar to 'matrix_row' that is being mapped to 'data' and vice versa, I could also map the db columns 'matrix_column' and 'matrix_value' to 'column' and 'value' respectively as well as let 'matrix_row' be 'row' mapping to 'data'. The names of the db columns don't change; it's just how I reference them in searches/selectors.

#1. This means that instead of the current prefixed sub-fields:

$pages->find("products.matrix_row=$foo, products.matrix_column=$bar, products.matrix_value=foo2");

#2. We'll rename the sub-fields as:

$pages->find("products.row=$foo, products.column=$bar, products.value=foo2");

Just looks cleaner and of course means less typing. Any strong objections to renaming the sub-fields this way (switching to #2 syntax above)? Been staring at this for too long! :-)

  • Like 1
Link to comment
Share on other sites

Version 1.0.0.

  • Changed status to stable
  • Updated version (big jump!) to 1 :-)
  • Changed db querying syntax slightly, i.e. subfields are now, e.g., products.row, products.column and products.value respectively for the matrix's rows, columns and values.

I have submitted this to the module's directory and also updated the first post in this thread.

  • Like 7
Link to comment
Share on other sites

Thanks @Macrura,

Btw, I corrected my 'cars' example in the first post. I had all cars in the same matrix table rather than car types and brands each in their own matrix (and page), allowing comparisons across cars (e.g. what type of car is best in xx) and within car types (e.g. in respect of Ford's, which brand does xx worst).

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...