Based on results posted, I decided to do some real world testing with a customer file with real data to see what kind of results I would get.
BACKGROUND: Instead of using val() to convert text numbers to values, I have my own library function called txt2num which performs the conversion but it also strips out punctuation such as dollar signs and commas as well as handles signs whether in front or behind and it deals with decimals that go beyond two places all on the fly.
FIRST TEST: Importing a small text file of 25 columns, 350 rows and 12 thisform.library.txt2num() conversion procedures per row that's 4,200 calls to the function.
1) Base run - No mdot references: 1.529 seconds 2) After mdot added - mdot references added to function: 1.436 seconds.
That works out to about a 6% improvement in speed. Not a huge improvement but it was small file.
SECOND TEST: Importing a larger text file of 50 columns, 10261 rows and 12 thisform.library.txt2num() conversion procedures per row or about 123,132 calls to the function.
1) Base run - No mdot references: 5.101 seconds 2) After mdot added - mdot references added to function: 4.399 seconds.
That works out to about a 13.7% improvement in speed.
Not sure if the improvement scales consistently as file sizes increase. However, it seems to me perhaps the most effective use of the mdot notation may be in refactoring it into libraries or classes. That way you will get the most advantage from functions you call repeatedly. Plus as soon as you re-compile an existing program you can gain the speed advantages immediately without refactoring ALL your code.
As always, your mileage may vary.
Paul H. Tarver Tarver Program Consultants, Inc. Email: paul@tpcqpc.com
-----Original Message----- From: ProfoxTech [mailto:profoxtech-bounces@leafe.com] On Behalf Of Darren Sent: Thursday, August 03, 2017 5:24 PM To: profoxtech@leafe.com Subject: RE: GETFILE() Returns Empty String in Some Cases
Personally I use this approach ...
STORE m.variable|obj.property|table.fieldname to variable. Fully qualify everything / always. Easier to debug as well.
Additional to results below ... STORE/ = | with/without m. | with/ without table open.
No Table Open x=y 0.1320 No Table Open STORE y to x 0.0870 No Table Open x=m.y 0.1328 No Table Open STORE m.y to x 0.0879 Table Open x=y 0.2081 Table Open STORE y to x 0.1644 Table Open x=m.y 0.1326 Table Open STORE m.y to x 0.0879 No Table Open STORE y to x 0.0870 No Table Open STORE m.y to x 0.0879 Table Open STORE m.y to x 0.0879 No Table Open x=y 0.1320 Table Open x=m.y 0.1326 No Table Open x=m.y 0.1328 Table Open STORE y to x 0.1644 Table Open x=y 0.2081
-----Original Message----- From: ProfoxTech [mailto:profoxtech-bounces@leafe.com] On Behalf Of Paul H. Tarver Sent: Wednesday, 2 August 2017 6:32 AM To: profoxtech@leafe.com Subject: RE: GETFILE() Returns Empty String in Some Cases
That was a really interesting article so I took the example, expanded it to include an object property and ran the test on my system.
Here are my results: --------------------------------- y=1 nsec=SECONDS() FOR i = 1 to 1000000 * x=y && Results: 0.377 * x=m.y && Results: 0.151 * x=fld255 && Results: 0.328 * x=test.fld255 && Results: 0.201 * x=objSession.nValue && Results: 0.377 ENDFOR ?SECONDS()-nsec ---------------------------------
Most of my work is done moving data through temporary cursors and I ALWAYS reference fields with the table.field method shown in the fourth test. I also often use object properties to store static variables. While it looks like I'm getting the best results using table.field access; I'm not getting any performance advantage using object properties or by not using the m. prefix. The good news is now I know if I'm going to do a LOT of iterations with memory variables, there are some performance gains to be picked up by simply using the m. notation. I think I may keep these results in mind when I'm doing refactoring on my function library!
This has been quite the learning experience and I really appreciate Ken for sharing that article! It also proves that no matter how long you've been programming in ANY language there is always something new you can pick up. Thanks!
Paul H. Tarver Tarver Program Consultants, Inc. Email: paul@tpcqpc.com
-----Original Message----- From: ProfoxTech [mailto:profoxtech-bounces@leafe.com] On Behalf Of Ken Dibble Sent: Tuesday, August 01, 2017 1:30 PM To: profoxtech@leafe.com Subject: RE: GETFILE() Returns Empty String in Some Cases
I remember!! It was Calvin Hsia! Here's the blog article:
https://blogs.msdn.microsoft.com/calvin_hsia/2004/12/14/foxpro-performance-t ip-field-name-lookup-for-tables/
Ken
On 2017-08-01 10:00, Ken Dibble wrote:
A member of the VFP Team (I can't remember his name; he was the guy who "took out the slow parts" from various subsystems for VFP 8 and 9) wrote a blog article explaining that without the M. prefix, the system MUST compare every variable referenced to every field in the current work area every time the variable is accessed to determine whether the variable is a field name. He said that if the current work area contains a large number of fields, and the same variable is used over a large number of iterations in a tight loop, using the M. prefix makes a considerable difference in speed. He showed test results to demonstrate this. I do a LOT of array iteration. So I began using the M. prefix for variables inside high-iteration tight loops. I noticed a slight speed increase. However, I find the M. prefix ugly and distracting. I used it yesterday as a quick fix to demonstrate what the problem was. I will probably eventually rename the variable. Ken
Very interesting...thanks for sharing that!
[excessive quoting removed by server]