I've written some very basic tools for grouping, pivoting, unioning and subtotaling datasets sourced from non DB sources (eg: CSV, OLTP systems). The "group by" methods sit at the core of most of these.
However i'm sure lot of work has been done in making efficient algorithms for grouping data... and i'm sure i'm not using them. And my Google-fu has completely failed to turn anything up.
Are there any good online sources or books describing the better methods to create grouped data?
Or should i just start looking at the MySQL source or something similar?
One very handy way to "group by" some field (or set of fields and expressions, but I'll use "field" for simplicity!-) is when you can arrange to walk over the results before grouping (RBG) in a sorted way -- you actually don't care about the sorting (save in the common case in which an ORDER BY is also there and just happens to be on the same field as the GROUP BY!-), but rather about the "side effect" property of ordering -- that all rows in RBG with the same value for the grouping field come right after each other, so you can accumulate until the grouping field changes, then emit/yield the results accumulated so far, and proceed to reinitialize the accumulators with the new row (the one with a different value of the grouping field) -- make sure to "just initialize the accumulators" at the very start, AND "just emit/yield accumulated results" at the very end, of course.
If this doesn't work, maybe you can hash the grouping field and use a hash table for the results being accumulated for that group -- at each row in RBG, hash the grouping field, check if it was already present as a key in the hash table, if not put it there with accumulators suitably initialized from the RBG row, else update the accumulators per the RBG row. You just emit everything at the end. The problem of course is you're taking up more memory until the end!-)
These are the two fundamental approaches. Would you like pseudocode for each, BTW?
You should check out OLAP databases. OLAP allows you to create a database of aggregates meant to be analyzed in a "slice and dice" fashion.
Aggregate measures such as counts, averages, mins, maxs, sums and stdev's can be quickly analyzed by any number of dimensions using an OLAP database.
See this introduction to OLAP on MSDN.
Give an example CSV file and type of result wanted and I might be able to rustle up a a solution in Python for you.
Python has the CSV module and list/generator comprehensions that can help with this sort of thing.
Paddy.
Related
Imagine a web form with a set of check boxes (any or all of them can be selected). I chose to save them in a comma separated list of values stored in one column of the database table.
Now, I know that the correct solution would be to create a second table and properly normalize the database. It was quicker to implement the easy solution, and I wanted to have a proof-of-concept of that application quickly and without having to spend too much time on it.
I thought the saved time and simpler code was worth it in my situation, is this a defensible design choice, or should I have normalized it from the start?
Some more context, this is a small internal application that essentially replaces an Excel file that was stored on a shared folder. I'm also asking because I'm thinking about cleaning up the program and make it more maintainable. There are some things in there I'm not entirely happy with, one of them is the topic of this question.
In addition to violating First Normal Form because of the repeating group of values stored in a single column, comma-separated lists have a lot of other more practical problems:
Can’t ensure that each value is the right data type: no way to prevent 1,2,3,banana,5
Can’t use foreign key constraints to link values to a lookup table; no way to enforce referential integrity.
Can’t enforce uniqueness: no way to prevent 1,2,3,3,3,5
Can’t delete a value from the list without fetching the whole list.
Can't store a list longer than what fits in the string column.
Hard to search for all entities with a given value in the list; you have to use an inefficient table-scan. May have to resort to regular expressions, for example in MySQL:
idlist REGEXP '[[:<:]]2[[:>:]]' or in MySQL 8.0: idlist REGEXP '\\b2\\b'
Hard to count elements in the list, or do other aggregate queries.
Hard to join the values to the lookup table they reference.
Hard to fetch the list in sorted order.
Hard to choose a separator that is guaranteed not to appear in the values
To solve these problems, you have to write tons of application code, reinventing functionality that the RDBMS already provides much more efficiently.
Comma-separated lists are wrong enough that I made this the first chapter in my book: SQL Antipatterns, Volume 1: Avoiding the Pitfalls of Database Programming.
There are times when you need to employ denormalization, but as #OMG Ponies mentions, these are exception cases. Any non-relational “optimization” benefits one type of query at the expense of other uses of the data, so be sure you know which of your queries need to be treated so specially that they deserve denormalization.
"One reason was laziness".
This rings alarm bells. The only reason you should do something like this is that you know how to do it "the right way" but you have come to the conclusion that there is a tangible reason not to do it that way.
Having said this: if the data you are choosing to store this way is data that you will never need to query by, then there may be a case for storing it in the way you have chosen.
(Some users would dispute the statement in my previous paragraph, saying that "you can never know what requirements will be added in the future". These users are either misguided or stating a religious conviction. Sometimes it is advantageous to work to the requirements you have before you.)
There are numerous questions on SO asking:
how to get a count of specific values from the comma separated list
how to get records that have only the same 2/3/etc specific value from that comma separated list
Another problem with the comma separated list is ensuring the values are consistent - storing text means the possibility of typos...
These are all symptoms of denormalized data, and highlight why you should always model for normalized data. Denormalization can be a query optimization, to be applied when the need actually presents itself.
In general anything can be defensible if it meets the requirements of your project. This doesn't mean that people will agree with or want to defend your decision...
In general, storing data in this way is suboptimal (e.g. harder to do efficient queries) and may cause maintenance issues if you modify the items in your form. Perhaps you could have found a middle ground and used an integer representing a set of bit flags instead?
Yes, I would say that it really is that bad. It's a defensible choice, but that doesn't make it correct or good.
It breaks first normal form.
A second criticism is that putting raw input results directly into a database, without any validation or binding at all, leaves you open to SQL injection attacks.
What you're calling laziness and lack of SQL knowledge is the stuff that neophytes are made of. I'd recommend taking the time to do it properly and view it as an opportunity to learn.
Or leave it as it is and learn the painful lesson of a SQL injection attack.
I needed a multi-value column, it could be implemented as an xml field
It could be converted to a comma delimited as necessary
querying an XML list in sql server using Xquery.
By being an xml field, some of the concerns can be addressed.
With CSV: Can't ensure that each value is the right data type: no way to prevent 1,2,3,banana,5
With XML: values in a tag can be forced to be the correct type
With CSV: Can't use foreign key constraints to link values to a lookup table; no way to enforce referential integrity.
With XML: still an issue
With CSV: Can't enforce uniqueness: no way to prevent 1,2,3,3,3,5
With XML: still an issue
With CSV: Can't delete a value from the list without fetching the whole list.
With XML: single items can be removed
With CSV: Hard to search for all entities with a given value in the list; you have to use an inefficient table-scan.
With XML: xml field can be indexed
With CSV: Hard to count elements in the list, or do other aggregate queries.**
With XML: not particularly hard
With CSV: Hard to join the values to the lookup table they reference.**
With XML: not particularly hard
With CSV: Hard to fetch the list in sorted order.
With XML: not particularly hard
With CSV: Storing integers as strings takes about twice as much space as storing binary integers.
With XML: storage is even worse than a csv
With CSV: Plus a lot of comma characters.
With XML: tags are used instead of commas
In short, using XML gets around some of the issues with delimited list AND can be converted to a delimited list as needed
Yes, it is that bad. My view is that if you don't like using relational databases then look for an alternative that suits you better, there are lots of interesting "NOSQL" projects out there with some really advanced features.
Well I've been using a key/value pair tab separated list in a NTEXT column in SQL Server for more than 4 years now and it works. You do lose the flexibility of making queries but on the other hand, if you have a library that persists/derpersists the key value pair then it's not a that bad idea.
I would probably take the middle ground: make each field in the CSV into a separate column in the database, but not worry much about normalization (at least for now). At some point, normalization might become interesting, but with all the data shoved into a single column you're gaining virtually no benefit from using a database at all. You need to separate the data into logical fields/columns/whatever you want to call them before you can manipulate it meaningfully at all.
If you have a fixed number of boolean fields, you could use a INT(1) NOT NULL (or BIT NOT NULL if it exists) or CHAR (0) (nullable) for each. You could also use a SET (I forget the exact syntax).
I have a system producing about 5TB of time-tagged numeric data every year. The fields tend to be different for each row, and to avoid having heaps of NULLs I'm thinking of using Postgres as a document store with JSONB.
However, GIN indexes on JSONB fields don't seem to be made for numerical and datetime data. There are no inequality or range operators for numbers and dates.
Here they suggest making special constructs with LATERAL to treat JSON values as normal numeric columns, and here someone proposes using a "sortable" string format for dates and filter string ranges.
These solutions sound a bit hacky and I wonder about their performance. Perhaps this is not a good application for JSONB?
An alternative I can think of using a relational DB is to use the 6th normal form, making one table for each (optional) field, of which however there would be hundreds. It sounds like a big JOIN mess, and new tables would have to be created on the fly any time a new field pops up. But maybe it's still better than a super-slow JSONB implementation.
Any guidance would be much appreciated.
More about the data
The data are mostly sensor readings, physical quantities and boolean flags. Which subset of these is present in each row is unpredictable. The index is an integer, and the only field that always exists is the corresponding date.
There would probably be one write for each value and almost no updates. Reads can be frequent and sliced based on any of the fields (some are more likely to be in a WHERE statement than others).
We have a database with hundreds of millions of records of log data. We're attempting to 'group' this log data as being likely to be of the same nature as other entries in the log database. For instance:
Record X may contain a log entry like:
Change Transaction ABC123 Assigned To Server US91
And Record Y may contain a log entry like:
Change Transaction XYZ789 Assigned To Server GB47
To us humans those two log entries are easily recognizable as being likely related in some way. Now, there may be 10 million rows between Record X and Record Y. And there may be thousands of other entries that are similar to X and Y, and some that are totally different but that have other records they are similar to.
What I'm trying to determine is the best way to group the similar items together and say that with XX% certainty Record X and Record Y are probably of the same nature. Or perhaps a better way of saying it would be that the system would look at Record Y and say based on your content you're most like Record X as apposed to all other records.
I've seen some mentions of Natural Language Processing and other ways to find similarity between strings (like just brute-forcing some Levenshtein calculations) - however for us we have these two additional challenges:
The content is machine generated - not human generated
As opposed to a search engine approach where we determine results for a given query - we're trying to classify a giant repository and group them by how alike they are to one another.
Thanks for your input!
Interesting problem. Obviously, there's a scale issue here because you don't really want to start comparing each record to every other record in the DB. I believe I'd look at growing a list of "known types" and scoring records against the types in that list to see if each record has a match in that list.
The "scoring" part will hopefully draw some good answers here -- your ability to score against known types is key to getting this to work well, and I have a feeling you're in a better position than we are to get that right. Some sort of soundex match, maybe? Or if you can figure out how to "discover" which parts of new records change, you could define your known types as regex expressions.
At that point, for each record, you can hopefully determine that you've got a match (with high confidence) or a match (with lower confidence) or very likely no match at all. In this last case, it's likely that you've found a new "type" that should be added to your "known types" list. If you keep track of the score for each record you matched, you could also go back for low-scoring matches and see if a better match showed up later in your processing.
I would suggest indexing your data using a text search engine like Lucene to split your log entries into terms. As your data is machine generated use also word bigrams and tigrams, even higher order n-grams. A bigram is just a sequence of consecutive words, in your example you would have the following bigrams:
Change_Transaction, Transaction_XYZ789, XYZ789_Assigned, Assigned_To, To_Server, Server_GB47
For each log prepare queries in a similar way, the search engine may give you the most similar results. You may need to tweek the similarity function a bit to obtain best results but I believe this is a good start.
Two main strategies come to my mind here:
the ad-hoc one. Use an information retrieval approach. Build an index for the log entries, eventually using a specialized tokenizer/parser, by feeding them into a regular text search engine. I've heard people do this with Xapian and Lucene. Then you can "search" for a new log record and the text search engine will (hopefully) return some related log entries to compare it with. Usually the "information retrieval" approach is however only interested in finding the 10 most similar results.
the clustering approach. You will usually need to turn the data into numerical vectors (that may however be sparse) e.g. as TF-IDF. Then you can apply a clustering algorithm to find groups of closely related lines (such as the example you gave above), and investigate their nature. You might need to tweak this a little, so it doesn't e.g. cluster on the server ID.
Both strategies have their ups and downs. The first one is quite fast, however it will always just return you some similar existing log lines, without much quantities on how common this line is. It's mostly useful for human inspection.
The second strategy is more computationally intensive, and depending on your parameters could fail completely (so maybe test it on a subset first), but could also give more useful results by actually building large groups of log entries that are very closely related.
It sounds like you could take the lucene approach mentioned above, then use that as a source for input vectors into the machine learning library Mahout (http://mahout.apache.org/). Once there you can train a classifier, or just use one of their clustering algorithms.
If your DBMS has it, take a look at SOUNDEX().
i am reading about the "entity attribute value model" which sort of reminds me of an star-schema which you use in data warehousing.
One table has all the facts (even if you mix apples,bananas e.g. date of farming, weight, price, color,type,name) and a bunch of tables holding the details (e.g. infected_with _banana_virus_type, apple_specific_acid_level)
I do this in both aproaches, so I can't see a difference in these to words?
Please enlighten me. CHEERS
In all approaches you have entities, attributes and values. Everything reduces to this logically. Since everything has entities, attributes and values, you can always claim that everything is the same. All data structures are -- from that point of view -- identical.
Please draw a diagram of a star schema. With a fact (say web site GET requests) and some dimensions like Time, IP Address, Requested Resource Path, and session User.
Actually draw the actual diagram, please. Don't read the words, look at the picture of five tables.
After drawing that picture, draw a single EAV table.
Actually draw the picture with entity, attribute and value columns. Don't read the words. Look at the picture of one table.
Okay?
Now write down all the differences between the two pictures. Number of tables. Number of columns. Data types of each column. All the differences.
We're not done.
Write a SQL query to count GET requests by day of the week for a given user using the star schema. Actually write the SQL. It's a three-table join. With a GROUP BY and a WHERE
Try and write a SQL query to count GET requests by day of week for the EAV table.
Okay?
Now write down all the differences between the two queries. Complexity of the SQL, for example. Performance of the SQL. Time required to write the SQL.
Now you know the differences.
I am trying to visualize how to create a search for an application that we are building. I would like a suggestion on how to approach 'searching' through large sets of data.
For instance, this particular search would be on a 750k record minimum table, of product sku's, sizing, material type, create date, etc;
Is anyone aware of a 'plugin' solution for Coldfusion to do this? I envision a google like single entry search where a customer can type in the part number, or the sizing, etc, and get hits on any or all relevant results.
Currently if I run a 'LIKE' comparison query, it seems to take ages (ok a few seconds, but still), and it is too long. At times making a user sit there and wait up to 10 seconds for queries & page loads.
Or are there any SQL formulas to help accomplish this? I want to use a proven method to search the data, not just a simple SQL like or = comparison operation.
So this is a multi-approach question, should I attack this at the SQL level (as it ultimately looks to be) or is there a plug in/module for ColdFusion that I can grab that will give me speedy, advanced search capability.
You could try indexing your db records with a Verity (or Solr, if CF9) search.
I'm not sure it would be faster, and whether even trying it would be worthwhile would depend a lot on how often you update the records you need to search. If you update them rarely, you could do an Verity Index update whenever you update them. If you update the records constantly, that's going to be a drag on the webserver, and certainly mitigate any possible gains in search speed.
I've never indexed a database via Verity, but I've indexed large collections of PDFs, Word Docs, etc, and I recall the search being pretty fast. I don't know if it will help your current situation, but it might be worth further research.
If your slowdown is specifically the search of textual fields (as I surmise from your mentioning of LIKE), the best solution is building an index table (not to be confiused with DB table indexes that are also part of the answer).
Build an index table mapping the unique ID of your records from main table to a set of words (1 word per row) of the textual field. If it matters, add the field of origin as a 3rd column in the index table, and if you want "relevance" features you may want to consider word count.
Populate the index table with either a trigger (using splitting) or from your app - the latter might be better, simply call a stored proc with both the actual data to insert/update and the list of words already split up.
This will immediately drastically speed up textual search as it will no longer do "LIKE", AND will be able to use indexes on index table (no pun intended) without interfering with indexing on SKU and the like on the main table.
Also, ensure that all the relevant fields are indexed fully - not necessarily in the same compund index (SKU, sizing etc...), and any field that is searched as a range field (sizing or date) is a good candidate for a clustered index (as long as the records are inserted in approximate order of that field's increase or you don't care about insert/update speed as much).
For anything mode detailed, you will need to post your table structure, existing indexes, the queries that are slow and the query plans you have now for those slow queries.
Another item is to enure that as little of the fields are textual as possible, especially ones that are "decodable" - your comment mentioned "is it boxed" in the text fields set. If so, I assume the values are "yes"/"no" or some other very limited data set. If so, simply store a numeric code for valid values and do en/de-coding in your app, and search by the numeric code. Not a tremendous speed improvement but still an improvement.
I've done this using SQL's full text indexes. This will require very application changes and no changes to the database schema except for the addition of the full text index.
First, add the Full Text index to the table. Include in the full text index all of the columns the search should perform against. I'd also recommend having the index auto update; this shouldn't be a problem unless your SQL Server is already being highly taxed.
Second, to do the actual search, you need to convert your query to use a full text search. The first step is to convert the search string into a full text search string. I do this by splitting the search string into words (using the Split method) and then building a search string formatted as:
"Word1*" AND "Word2*" AND "Word3*"
The double-quotes are critical; they tell the full text index where the words begin and end.
Next, to actually execute the full text search, use the ContainsTable command in your query:
SELECT *
from containstable(Bugs, *, '"Word1*" AND "Word2*" AND "Word3*"')
This will return two columns:
Key - The column identified as the primary key of the full text search
Rank - A relative rank of the match (1 - 1000 with a higher ranking meaning a better match).
I've used approaches similar to this many times and I've had good luck with it.
If you want a truly plug-in solution then you should just go with Google itself. It sounds like your doing some kind of e-commerce or commercial site (given the use of the term 'SKU'), So you probably have a catalog of some kind with product pages. If you have consistent markup then you can configure a google appliance or service to do exactly what you want. It will send a bot in to index your pages and find your fields. No SQl, little coding, it will not be dependent on your database, or even coldfusion. It will also be quite fast and familiar to customers.
I was able to do this with a coldfusion site in about 6 hours, done! The only thing to watch out for is that google's index is limited to what the bot can see, so if you have a situation where you want to limit access based on a users role or permissions or group, then it may not be the solution for you (although you can configure a permission service for Google to check with)
Because SQL Server is where your data is that is where your search performance is going to be a possible issue. Make sure you have indexes on the columns you are searching on and if using a like you can't use and index if you do this SELECT * FROM TABLEX WHERE last_name LIKE '%FR%'
But it can use an index if you do it like this SELECT * FROM TABLEX WHERE last_name LIKE 'FR%'. The key here is to allow as many of the first characters to not be wild cards.
Here is a link to a site with some general tips. https://web.archive.org/web/1/http://blogs.techrepublic%2ecom%2ecom/datacenter/?p=173