Revision control for multiple pieces of related data - php

I'm trying to figure out how to best keep revision/history information on revisions to multiple rows of data, in case for some reason we need to revert to that data.
This is the general sort of layout:
item
---------------
id
title
etc...
region
---------------
id
title
etc...
release_type
-----------------
id
title
etc...
items_released_dates_data
---------------------
item_id
region_id
release_type_id (these three form the primary key)
date
So you can have one release date per item + region_id + release_type and we basically only track the date (For the purposes of this question the 'date' could be a number, a string, or whatever. I'm certain to run into this issue again)
Changes are submitted in bulk, when new data is added everything in items_released_dates_data where item_id=your_id is first deleted then an insert statement adds the new values (perhaps this isn't the best way to do this?)
My thought was to create a table like:
items_release_dates_data_history
-------------------------------------
item_id
timestamp
description
raw_data
Making description a short summary of what was updated, and including the data in some format like json or xml or something that could be quickly decoded on the client side to give the user a review of the changes and a choice to revise to a given version. Then every entry to items_released_dates_data also requires an entry to items_released_dates_data_history (doesn't sound like a question does it? :| )
I've read something about mysql triggers that would be helpful here, but quite frankly I don't know a thing about them so I'm working with what I understand.
My question is, am I following the right path to version this stuff, and is there any advice/best practices anyone can give me on how to improve this method?

I second Alex Miller's comment. Everything you write make sense so far.
I'd strongly recommend looking into triggers though, despite your reservations. They're fairly easy to grasp, and make for a very powerful tool in such scenarios. Using triggers you can store a copy of the row into a separate table each time a record is updated (or deleted). If you want to go all fancy you can, within the trigger, compare the incoming data to the existing data, and write only what has changed.
Also consider the Archive storage engine instead of MyISAM or InnoDB for these kinds of tables - they're made for this kind of job.
Also, the search phrase you're probably looking for is "audit trail".

I'd say that you're definitely on the right track. Although, you may want to store the region ID in the history so you can check release history based on a region rather than just by entire items.
As for the delete+insert, that's fine as long as you don't end up with too much traffic, as those are both locking actions. There is a lot of time used when inserting or deleting a row to update the index. If you're using a MyISAM table, it's also going to halt all reads on the table until those actions complete. Update will as well, but for a much shorter time. InnoDB will only lock the row, so that's not really a concern.

Related

SQL - auto increment withing group inside one table [duplicate]

I have got a table which has an id (primary key with auto increment), uid (key refering to users id for example) and something else which for my question won’t matter.
I want to make, lets call it, different auto-increment keys on id for each uid entry.
So, I will add an entry with uid 10, and the id field for this entry will have a 1 because there were no previous entries with a value of 10 in uid. I will add a new one with uid 4 and its id will be 3 because I there were already two entried with uid 4.
...Very obvious explanation, but I am trying to be as explainative an clear as I can to demonstrate the idea... clearly.
What SQL engine can provide such a functionality natively? (non Microsoft/Oracle based)
If there is none, how could I best replicate it? Triggers perhaps?
Does this functionality have a more suitable name?
In case you know about a non SQL database engine providing such a functioality, name it anyway, I am curious.
Thanks.
MySQL's MyISAM engine can do this. See their manual, in section Using AUTO_INCREMENT:
For MyISAM tables you can specify AUTO_INCREMENT on a secondary column in a multiple-column index. In this case, the generated value for the AUTO_INCREMENT column is calculated as MAX(auto_increment_column) + 1 WHERE prefix=given-prefix. This is useful when you want to put data into ordered groups.
The docs go on after that paragraph, showing an example.
The InnoDB engine in MySQL does not support this feature, which is unfortunate because it's better to use InnoDB in almost all cases.
You can't emulate this behavior using triggers (or any SQL statements limited to transaction scope) without locking tables on INSERT. Consider this sequence of actions:
Mario starts transaction and inserts a new row for user 4.
Bill starts transaction and inserts a new row for user 4.
Mario's session fires a trigger to computes MAX(id)+1 for user 4. You get 3.
Bill's session fires a trigger to compute MAX(id). I get 3.
Bill's session finishes his INSERT and commits.
Mario's session tries to finish his INSERT, but the row with (userid=4, id=3) now exists, so Mario gets a primary key conflict.
In general, you can't control the order of execution of these steps without some kind of synchronization.
The solutions to this are either:
Get an exclusive table lock. Before trying an INSERT, lock the table. This is necessary to prevent concurrent INSERTs from creating a race condition like in the example above. It's necessary to lock the whole table, since you're trying to restrict INSERT there's no specific row to lock (if you were trying to govern access to a given row with UPDATE, you could lock just the specific row). But locking the table causes access to the table to become serial, which limits your throughput.
Do it outside transaction scope. Generate the id number in a way that won't be hidden from two concurrent transactions. By the way, this is what AUTO_INCREMENT does. Two concurrent sessions will each get a unique id value, regardless of their order of execution or order of commit. But tracking the last generated id per userid requires access to the database, or a duplicate data store. For example, a memcached key per userid, which can be incremented atomically.
It's relatively easy to ensure that inserts get unique values. But it's hard to ensure they will get consecutive ordinal values. Also consider:
What happens if you INSERT in a transaction but then roll back? You've allocated id value 3 in that transaction, and then I allocated value 4, so if you roll back and I commit, now there's a gap.
What happens if an INSERT fails because of other constraints on the table (e.g. another column is NOT NULL)? You could get gaps this way too.
If you ever DELETE a row, do you need to renumber all the following rows for the same userid? What does that do to your memcached entries if you use that solution?
SQL Server should allow you to do this. If you can't implement this using a computed column (probably not - there are some restrictions), surely you can implement it in a trigger.
MySQL also would allow you to implement this via triggers.
In a comment you ask the question about efficiency. Unless you are dealing with extreme volumes, storing an 8 byte DATETIME isn't much of an overhead compared to using, for example, a 4 byte INT.
It also massively simplifies your data inserts, as well as being able to cope with records being deleted without creating 'holes' in your sequence.
If you DO need this, be careful with the field names. If you have uid and id in a table, I'd expect id to be unique in that table, and uid to refer to something else. Perhaps, instead, use the field names property_id and amendment_id.
In terms of implementation, there are generally two options.
1). A trigger
Implementations vary, but the logic remains the same. As you don't specify an RDBMS (other than NOT MS/Oracle) the general logic is simple...
Start a transaction (often this is Implicitly already started inside triggers)
Find the MAX(amendment_id) for the property_id being inserted
Update the newly inserted value with MAX(amendment_id) + 1
Commit the transaction
Things to be aware of are...
- multiple records being inserted at the same time
- records being inserted with amendment_id being already populated
- updates altering existing records
2). A Stored Procedure
If you use a stored procedure to control writes to the table, you gain a lot more control.
Implicitly, you know you're only dealing with one record.
You simply don't provide a parameter for DEFAULT fields.
You know what updates / deletes can and can't happen.
You can implement all the business logic you like without hidden triggers
I personally recommend the Stored Procedure route, but triggers do work.
It is important to get your data types right.
What you are describing is a multi-part key. So use a multi-part key. Don't try to encode everything into a magic integer, you will poison the rest of your code.
If a record is identified by (entity_id,version_number) then embrace that description and use it directly instead of mangling the meaning of your keys. You will have to write queries which constrain the version number but that's OK. Databases are good at this sort of thing.
version_number could be a timestamp, as a_horse_with_no_name suggests. This is quite a good idea. There is no meaningful performance disadvantage to using timestamps instead of plain integers. What you gain is meaning, which is more important.
You could maintain a "latest version" table which contains, for each entity_id, only the record with the most-recent version_number. This will be more work for you, so only do it if you really need the performance.

mysql historical data and record id

I am setting up a new part of an application with historical data requirements for the transactions table in mysql. Originally in old version transactions were not historical, with structure like this:
id|buyerid|prodid|price|status
And other fields, with the id being referenced in links to access Transaction Details page, as well as used as foreign key in other tables across the application to reference particular transactions for various purposes.
Now the requirement is to answer reporting questions like "Show all transaction that had particular status Feb 2014" AND "What did a transaction look like in Feb 2014".
The new design I'm testing at the moment is below:
id|buyerid|prodid|price|status|active|start_date|end_date
Where active used to indicate latest record, start is when it is created, no records to be modified instead end date populated and a new record created with same details plus the modification.
Now the question is - what to do about transaction id field? Because in this new design it is more of a history id, and can not be used for a foreign key across the application since it is going to change with every update.
I can think of two options:
Create a separate table, transaction_ids with just one column, primary key autoincrement tid, and a foreign key column in the main transactions table for tid - Every time a brand new transaction is created, insert the ids table and use that id for the tid to trace this particular transaction across the system.
The buyerid and prodid combination is always unique in my application, no buyer can get the same product twice.
Is the second solution better? Does anyone know of a better way to handle this?
What you are trying to achieve is called Event Sourcing.
Think in terms of events changing the status of your transaction, rather than tracing the status itself in time.
You still have your transaction with its own primary key, and you rebuild the current (or past) status applying each event.
I would also suggest you to start coding your business models, and only after that, to think about the persistence and the best way to map it to a database.
Second Solution looks better although I will say that there is a lot of ambiguity in your question.
I am saying that second solution is better because the transaction_ids table which you are talking about in solution 1 is basically REDUNDANT. It is not solving any purpose. Even if the transaction id is repeating itself in the transaction table, it does not mean that you need to have a separate table to generate the ids and make it as PK-FK relation. Most probably you will still be querying the data by user-id and prod-id and not by transaction-id
Basically what you need is some kind of audit history table where you insert a record for every operation/transaction/modification done and capture some basic details like - Username, Date/time, old value, new value etc. You do not need status or start date and end date columns. Once a record is inserted in this audit history table then it is never going to be touched again.
You will have to design your report carefully.
Taking two previous answers into consideration, here is the solution I will go with: All of the data updates in my application come through one single function, that is already set up to audit particular fields of my choosing, so I will mark the transaction status to be audited among the others. Table structure for the audit table is similar to this:
|id|table|table_id|column|old_val|new_val|who|when|
Only that there is a bit more advanced object mapping via object id's instead of simple table name. I can then use this data in a Join to the main, normal not historical transactions table to provide the reporting required.

Soft delete best practices (PHP/MySQL)

Problem
In a web application dealing with products and orders, I want to maintain information and relationships between former employees (users) and the orders they handled. I want to maintain information and relationships between obsolete products and orders which include these products.
However I want employees to be able to de-clutter the administration interfaces, such as removing former employees, obsolete products, obsolete product groups etc.
I'm thinking of implementing soft-deletion. So, how does one usually do this?
My immediate thoughts
My first thought is to stick a "flag_softdeleted TINYINT NOT NULL DEFAULT 0" column in every table of objects that should be soft deletable. Or maybe use a timestamp instead?
Then, I provide a "Show deleted" or "Undelete" button in each relevant GUI. Clicking this button you will include soft-deleted records in the result. Each deleted record has a "Restore" button. Does this make sense?
Your thoughts?
Also, I'd appreciate any links to relevant resources.
That's how I do it. I have a is_deleted field which defaults to 0. Then queries just check WHERE is_deleted = 0.
I try to stay away from any hard-deletes as much as possible. They are necessary sometimes, but I make that an admin-only feature. That way we can hard-delete, but users can't...
Edit: In fact, you could use this to have multiple "layers" of soft-deletion in your app. So each could be a code:
0 -> Not Deleted
1 -> Soft Deleted, shows up in lists of deleted items for management users
2 -> Soft Deleted, does not show up for any user except admin users
3 -> Only shows up for developers.
Having the other 2 levels will still allow managers and admins to clean up the deleted lists if they get too long. And since the front-end code just checks for is_deleted = 0, it's transparent to the frontend...
Using soft-deletes is a common thing to implement, and they are dead useful for lots of things, like:
Saving a user's data when they deleted something
Saving your own data when you delete something
Keep a track record of what really happened (a kind of audit)
etcetera
There is one thing I want to point out that almost everyone miss, and it always comes back to bite you in the rear piece. The users of your application does not have the same understanding of a delete as you have.
There are different degrees of deletions. The typical user deletes stuff when (s)he
Made a misstake and want to remove the bad data
Doesn't want to see something on the screen anymore
The problem is that if you don't record the intention of the delete, your application cannot distinguish between erronous data (that should never have been created) and historically correct data.
Have a look at the following data:
PRICES | item | price | deleted |
+------+-------+---------+
| A | 101 | 1 |
| B | 110 | 1 |
| C | 120 | 0 |
+------+-------+---------+
Some user doesn't want to show the price of item B, since they don't sell that item anymore. So he deletes it. Another user created a price for item A by misstake, so he deleted it and created the price for item C, as intended. Now, can you show me a list of the prices for all products? No, because either you have to display potentially erronous data (A), or you have to exclude all but current prices (C).
Of course the above can be dealt with in any number of ways. My point is that YOU need to be very clear with what YOU mean by a delete, and make sure that there is no way for the users to missunderstand it. One way would be to force the user to make a choice (hide/delete).
If I had existing code that hits that table, I would add the column and change the name of the table. Then I would create a view with the same name as the current table which selects only the active records. That way none of the existing code woudl break and you could have the soft delete column. If you want to see the deleted record, you select from the base table, otherwise you use the view.
I've always just used a deleted column as you mentioned. There's really not much more to it than that. Instead of deleting the record, just set the deleted field to true.
Some components I build allow the user to view all deleted records and restore them, others just display all records where deleted = 0
Your idea does make sense and is used frequently in production but, to implement it you will need to update quite a bit of code to account for the new field. Another option could be to archive (move) the "soft-deleted" records to a separate table or database. This is done frequently as well and makes the issue one of maintenance rather than (re)programming. (You could have a table trigger react to the delete to archive the deleted record.)
I would do the archiving to avoid a major update to production code. But if you want to use deleted-flag field, use it as a timestamp to give you additional useful info beyond a boolean. (Null = not deleted.) You might also want to add a DeletedBy field to track the user responsible for deleting the record. Using two fields gives you a lot of info tells you who deleted what and when. (The two extra field solution is also something that can be done in an archive table/database.)
The most common scenario I've come across is what you describe, a tinyint or even bit representing a status of IsActive or IsDeleted. Depending on whether this is considered "business" or "persistence" data it may be baked into the application/domain logic as transparently as possible, such as directly in stored procedures and not known to the application code. But it sounds like this is legitimate business information for your needs so would need to be known throughout the code. (So users can view deleted records, as you suggest.)
Another approach I've seen is to use a combination of two timestamps to show a "window" of activity for a given record. It's a little more code to maintain it, but the benefit is that something can be scheduled to soft-delete itself at a pre-determined time. Limited-time products can be set that way when they're created, for example. (To make a record active indefinitely one could use a max value (or just some absurdly distant future date) or just have the end date be null if you're ok with that.)
Then of course there's further consideration of things being deleted/undeleted from time to time and tracking some kind of audit for that. The flag approach knows only the current status, the timestamp approach knows only the most recent window. But anything as complex as an audit trail should definitely be stored separately than the records in question.
Instead I would use a bin table in which to move all the records deleted from the other tables. The main problem with the delete flag is that with linked tables you will definitely run into a double key error when trying to insert a new record.
The bin table could have a structure like this:
id, table_name, data, date_time, user
Where
id is the primary key with auto increment
table_name is the name of the table from which the record was deleted
data contains the record in JSON format with name and value of all fields
date_time is the date and time of the deletion
user is the identifier of the user (if the system provides for it) who performed the operation
this method will not only save you from checking the delete flag at each query (immagine the ones with many joins), but will allow you to have only the really necessary data in the tables, facilitating any searches and corrections using SQL client programs

How to design a generic database whose layout may change over time?

Here's a tricky one - how do I programatically create and interrogate a database whose contents I can't really foresee?
I am implementing a generic input form system. The user can create PHP forms with a WYSIWYG layout and use them for any purpose he wishes. He can also query the input.
So, we have three stages:
a form is designed and generated. This is a one-off procedure, although the form can be edited later. This designs the database.
someone or several people make use of the form - say for daily sales reports, stock keeping, payroll, etc. Their input to the forms is written to the database.
others, maybe management, can query the database and generate reports.
Since these forms are generic, I can't predict the database structure - other than to say that it will reflect HTML form fields and consist of a the data input from collection of edit boxes, memos, radio buttons and the like.
Questions and remarks:
A) how can I best structure the database, in terms of tables and columns? What about primary keys? My first thought was to use the control name to identify each column, then I realized that the user can edit the form and rename, so that maybe "name" becomes "employee" or "wages" becomes ":salary". I am leaning towards a unique number for each.
B) how best to key the rows? I was thinking of a timestamp to allow me to query and a column for the row Id from A)
C) I have to handle column rename/insert/delete. Foe deletion, I am unsure whether to delete the data from the database. Even if the user is not inputting it from the form any more he may wish to query what was previously entered. Or there may be some legal requirements to retain the data. Any gotchas in column rename/insert/delete?
D) For the querying, I can have my PHP interrogate the database to get column names and generate a form with a list where each entry has a database column name, a checkbox to say if it should be used in the query and, based on column type, some selection criteria. That ought to be enough to build searches like "position = 'senior salesman' and salary > 50k".
E) I probably have to generate some fancy charts - graphs, histograms, pie charts, etc for query results of numerical data over time. I need to find some good FOSS PHP for this.
F) What else have I forgotten?
This all seems very tricky to me, but I am database n00b - maybe it is simple to you gurus?
Edit: please don't tell me not to do it. I don't have any choice :-(
Edit: in real life I don't expect column rename/insert/delete to be frequent. However it is possible that after running for a few months a change to the database might be required. I am sure this happens regularly. I fear that I have worded this question badly and that people think that changes will be made willy-nilly every 10 minutes or so.
Realistically, my users will define a database when they lay out the form. They might get it right first time and never change it - especially if they are converting from paper forms. Even if they do decide to change, this might only happen once or twice ever, after months or years - and that can happen in any database.
I don't think that I have a special case here, nor that we should be concentrating on change. Perhaps better to concentrate on linkage - what's a good primary key scheme? Say, perhaps, for one text input, one numerical and a memo?
"This all seems very tricky to me, but
I am database n00b - maybe it is
simple to you gurus?"
Nope, it really is tricky. Fundamentally what you're describing is not a database application, it is a database application builder. In fact, it sounds as if you want to code something like Google App Engine or a web version of MS Access. Writing such a tool will take a lot of time and expertise.
Google has implemented flexible schemas by using its BigTable platform. It allows you to flex the schema pretty much at will. The catch is, this flexibility makes it very hard to write queries like "position = 'senior salesman' and salary > 50k".
So I don't think the NoSQL approach is what you need. You want to build an application which generates and maintains RDBMS schemas. This means you need to design a metadata repository from which you can generate dynamic SQL to build and change the users' schemas and also generate the front end.
Things your metadata schema needs to store
For schema generation:
foreign key relationships (an EMPLOYEE works in a DEPARTMENT)
unique business keys (there can be only one DEPARTMENT called "Sales")
reference data (permitted values of EMPLOYEE.POSITION)
column data type, size, etc
whether column is optional (i.e NULL or NOT NULL)
complex business rules (employee bonuses cannot exceed 15% of their salary)
default value for columns
For front-end generation
display names or labels ("Wages", "Salary")
widget (drop down list, pop-up calendar)
hidden fields
derived fields
help text, tips
client-side validation (associated JavaScript, etc)
That last points to the potential complexity in your proposal: a regular form designer like Joe Soap is not going to be able to formulate the JS to (say) validate that an input value is between X and Y, so you're going to have to derive it using templated rules.
These are by no means exhaustive lists, it's just off the top of my head.
For primary keys I suggest you use a column of GUID datatype. Timestamps aren't guaranteed to be unique, although if you run your database on an OS which goes to six places (i.e. not Windows) it's unlikely you'll get clashes.
last word
'My first thought was to use the
control name to identify each column,
then I realized that the user can edit
the form and rename, so that maybe
"name" becomes "employee" or "wages"
becomes ":salary". I am leaning
towards a unique number for each.'
I have built database schema generators before. They are hard going. One thing which can be tough is debugging the dynamic SQL. So make it easier on yourself: use real names for tables and columns. Just because the app user now wants to see a form titled HEADCOUNT it doesn't mean you have to rename the EMPLOYEES table. Hence the need to separate the displayed label from the schema object name. Otherwise you'll find yourself trying to figure out why this generated SQL statement failed:
update table_11123
set col_55542 = 'HERRING'
where col_55569 = 'Bootle'
/
That way madness lies.
In essence, you are asking how to build an application without specifications. Relational databases were not designed so that you can do this effectively. The common approach to this problem is an Entity-Attribute-Value design and for the type of system in which you want to use it, the odds of failure are nearly 100%.
It makes no sense for example, that the column called "Name" could become "Salary". How would a report where you want the total salary work if the salary values could have "Fred", "Bob", 100K, 1000, "a lot"? Databases were not designed to let anyone put anything anywhere. Successful database schemas require structure which means effort with respect to specifications on what needs to be stored and why.
Therefore, to answer your question, I would rethink the problem. The entire approach of trying to make an app that can store anything in the universe is not a recipe for success.
Like Thomas said, rational database is not good at your problem. However, you may want to take a look at NoSQL dbs like MongoDB.
See this article:
http://www.simple-talk.com/opinion/opinion-pieces/bad-carma/
for someone else's experience of your problem.
This is for A) & B), and is not something I have done but thought it was an interesting idea that Reddit put to use, see this link (look at Lesson 3):
http://highscalability.com/blog/2010/5/17/7-lessons-learned-while-building-reddit-to-270-million-page.html
Not sure about the database but for charts instead of using PHP for the charts, I recommend looking into using javascript (http://www.reynoldsftw.com/2009/02/6-jquery-chart-plugins-reviewed/). Advantages to this are some of the processing is offloaded to the client side for chart displays and they can be interactive.
The other respondents are correct that you should be very cautious with this approach because it is more complex and less performant than the traditional relational model - but I've done this type of thing to accommodate departmental differences at work, and it worked fine for the amount of use it got.
Basically I set it up like this, first - a table to store some information about the Form the user wants to create (obviously, adjust as you need):
--************************************************************************
-- Create the User_forms table
--************************************************************************
create table User_forms
(
form_id integer identity,
name varchar(200),
status varchar(1),
author varchar(50),
last_modifiedby varchar(50),
create_date datetime,
modified_date datetime
)
Then a table to define the fields to be presented on the form including any limits
and the order and page they are to be presented (my app presented the fields as a
multi-page wizard type of flow).
-
-************************************************************************
-- Create the field configuration table to hold the entry field configuration
--************************************************************************
create table field_configuration
(
field_id integer identity,
form_id SMALLINT,
status varchar(1),
fieldgroup varchar(20),
fieldpage integer,
fieldseq integer,
fieldname varchar(40),
fieldwidth integer,
description varchar(50),
minlength integer,
maxlength integer,
maxval varchar(13),
minval varchar(13),
valid_varchars varchar(20),
empty_ok varchar(1),
all_caps varchar(1),
value_list varchar(200),
ddl_queryfile varchar(100),
allownewentry varchar(1),
query_params varchar(50),
value_default varchar(20)
);
Then my perl code would loop through the fields in order for page 1 and put them on the "wizard form" ... and the "next" button would present the page 2 fields in order etc.
I had javascript functions to enforce the limits specified for each field as well ...
Then a table to hold the values entered by the users:
--************************************************************************
-- Field to contain the values
--************************************************************************
create table form_field_values
(
session_Id integer identity,
form_id integer,
field_id integer,
value varchar(MAX)
);
That would be a good starting point for what you want to do, but keep an eye on performance as it can really slow down any reports if they add 1000 custom fields. :-)

mysql show table / columns - performance question

I'm working on a basic php/mysql CMS and have a few questions regarding performance.
When viewing a blog page (or other sortable data) from the front-end, I want to allow a simple 'sort' variable to be added to the querystring, allowing posts to be sorted by any column. Obviously I can't accept anything from the querystring, and need to make sure the column exists on the table.
At the moment I'm using
SHOW TABLES;
to get a list of all of the tables in the database, then looping the array of table names and performing
SHOW COLUMNS;
on each.
My worry is that my CMS might take a performance hit here. I thought about using a static array of the table names but need to keep this flexible as I'm implementing a plugin system.
Does anybody have any suggestions on how I can keep this more concise?
Thankyou
If you using mysql 5+ then you'll find database information_schema usefull for your task. In this database you can access information of tables, columns, references by simple SQL queries. For example you can find if there is specific column at the table:
SELECT count(*) from COLUMNS
WHERE
TABLE_SCHEMA='your_database_name' AND
TABLE_NAME='your_table' AND
COLUMN_NAME='your_column';
Here is list of tables with specific column exists:
SELECT TABLE_SCHEMA, TABLE_NAME from COLUMNS WHERE COLUMN_NAME='your_column';
Since you're currently hitting the db twice before you do your actual query, you might want to consider just wrapping the actual query in a try{} block. Then if the query works you've only done one operation instead of 3. And if the query fails, you've still only wasted one query instead of potentially two.
The important caveat (as usual!) is that any user input be cleaned before doing this.
You could query the table up front and store the columns in a cache layer (i.e. memcache or APC). You could then set the expire time on the file to infinite and only delete and re-create the cache file when a plugin has been newly added, updated, etc.
I guess the best bet is to put all that stuff ur getting from Show tables etc in a file already and just include it, instead of running that every time. Or implement some sort of caching if the project is still in development and u think the fields will change.

Categories