While the general opinion of the Haskell community seems to be that it's always better to use Text instead of String, the fact that still the APIs of most of maintained libraries are String-oriented confuses the hell out of me. On the other hand, there are notable projects, which consider String as a mistake altogether and provide a Prelude with all instances of String-oriented functions replaced with their Text-counterparts.
So are there any reasons for people to keep writing String-oriented APIs except backwards- and standard Prelude-compatibility and the "switch-making inertia"?
Are there possibly any other drawbacks to Text as compared to String?
Particularly, I'm interested in this because I'm designing a library and trying to decide which type to use to express error messages.
My unqualified guess is that most library writers don't want to add more dependencies than necessary. Since strings are part of literally every Haskell distribution (it's part of the language standard!), it is a lot easier to get adopted if you use strings and don't require your users to sort out Text distributions from hackage.
It's one of those "design mistakes" that you just have to live with unless you can convince most of the community to switch over night. Just look at how long it has taken to get Applicative to be a superclass of Monad – a relatively minor but much wanted change – and imagine how long it would take to replace all the String things with Text.
To answer your more specific question: I would go with String unless you get noticeable performance benefits by using Text. Error messages are usually rather small one-off things so it shouldn't be a big problem to use String.
On the other hand, if you are the kind of ideological purist that eschews pragmatism for idealism, go with Text.
* I put design mistakes in scare quotes because strings as a list-of-chars is a neat property that makes them easy to reason about and integrate with other existing list-operating functions.
If your API is targeted at processing large amounts of character oriented data and/or various encodings, then your API should use Text.
If your API is primarily for dealing with small one-off strings, then using the built-in String type should be fine.
Using String for large amounts of text will make applications using your API consume significantly more memory. Using it with foreign encodings could seriously complicate usage depending on how your API works.
String is quite expensive (at least 5N words where N is the number of Char in the String). A word is same number of bits as the processor architecture (ex. 32 bits or 64 bits):
http://blog.johantibell.com/2011/06/memory-footprints-of-some-common-data.html
There are at least three reasons to use [Char] in small projects.
[Char] does not rely on any arcane staff, like foreign pointers, raw memory, raw arrays, etc that may work differently on different platforms or even be unavailable altogether
[Char] is the lingua franka in haskell. There are at least three 'efficient' ways to handle unicode data in haskell: utf8-bytestring, Data.Text.Text and Data.Vector.Unboxed.Vector Char, each requiring dealing with extra package.
by using [Char] one gains access to all power of [] monad, including many specific functions (alternative string packages do try to help with it, but still)
Personally, I consider utf16-based Data.Text one of the most questionable desicions of the haskell community, since utf16 combines flaws of both utf8 and utf32 encoding while having none of their benefits.
I wonder if Data.Text is always more efficient than Data.String???
"cons" for instance is O(1) for Strings and O(n) for Text. Append is O(n) for Strings and O(n+m) for strict Text's. Likewise,
let foo = "foo" ++ bigchunk
bar = "bar" ++ bigchunk
is more space efficient for Strings than for strict Texts.
Other issue not related to efficiency is pattern matching (perspicuous code) and lazyness (predictably per-character in Strings, somehow implementation dependent in lazy Text).
Text's are obviously good for static character sequences and for in-place modification. For other forms of structural editing, Data.String might have advantages.
I do not think there is a single technical reason for String to remain.
And I can see several ones for it to go.
Overall I would first argue that in the Text/String case there is only one best solution :
String performances are bad, everyone agrees on that
Text is not difficult to use. All functions commonly used on String are available on Text, plus some useful more in the context of strings (substitution, padding, encoding)
having two solutions creates unnecessary complexity unless all base functions are made polymorphic. Proof : there are SO questions on the subject of automatic conversions. So this is a problem.
So one solution is less complex than two, and the shortcomings of String will make it disappear eventually. The sooner the better !
Related
I'm working on a large application that constructs a lot of Data.Text values on the fly. I've been building all my Text values using (<>) and Data.Text.concat.
I only recently learned of the existence of the Builder type. The Beginning Haskell book has this to say about it:
Every time two elements are concatenated, a new Text value has to be created, and this comes with some overhead to allocate memory, to copy data, and also to keep track of the value and release it when it's no longer needed... Both the text and bytestring packages provide a Builder data type that can be used to efficiently generate large text values. [pg 240]
However, the book doesn't give any indication of exactly what is meant by "large text values."
So, I'm wondering whether or not I should refactor my code to use Builder. Maybe you can help me make that decision.
Specifically, I have these questions:
1) Are there any guidelines or "best practices" regarding when one should choose Builder over concatenation? Or, how do I know that a given Text value is "large" enough that it merits using Builder?
2) Is using Builder a "no brainer," or would it be worthwhile doing some profiling to confirm its benefits before undertaking a large-scale refactoring?
Thanks!
Data.Text.concat is an O(n+m) operation where n and m are the lengths of the strings you want to concat. This is because a new memory buffer of size n + m must be allocated to store the result of the concatenation.
Builder is specifically optimized for the mappend operation. It's a cheap O(1) operation (function composition, which is also excellently optimized by GHC). With Builder you are essentially building up the instructions for how to produce the final string result, but delaying the actual creation until you do some Builder -> Text transformation.
To answer your questions, you should choose Builder if you have profiled your application and discovered that Text.concat are dominating the run time. This will obviously depend on your needs and application. There is no general rule for when you should use Builder but for short Text literals there is probably no need.
Profiling would definitely be worthwhile if using Builder would involve "undertaking a large-scale refactoring". Although it goes without saying that Haskell will naturally make this kind of refactoring much less painful than you might be used to with less developer friendly languages, so it might not be such a difficult undertaking after all.
Do you know the fastest way to encode and decode UTF8 with some extra information? Here's the interesting cases that occur to me:
Serialization
I just want to encode an opaque buffer with no validation so I can decode again later. The fastest would be to use the underlying memory buffer and somehow unsafely coerce it from Text to ByteString without touching the contents.
Probably ASCII
I guess that 99% of the time my UTF8 is actually ASCII so it makes sense to do a first pass to confirm this and only further processing if it's found not to be true.
Probably not ASCII
Converse of the previous.
Probably short
A single key in JSON or a database that I guess will be 1 to 20 characters. Would be silly pay some upfront cost like vectorized SIMD approach.
Probably long
An HTML document. It's worth it pay some upfront cost for the highest throughput.
There's some more variants that are similar like if encoding JSON or URL and you think there's probably no escape characters.
I'm asking this question under the [Haskell] tag since Haskell's strong typing makes some techniques that would be easy in, say, C hard to implement. Also, there may be some special GHC tricks like using SSE4 instructions on an Intel platform that would be interesting. But this is more of a UTF8 issue in general and good ideas would be helpful to any language.
Update
After some research I propose to implement encode and decode for serialization purposes like so:
myEncode :: Text -> ByteString
myEncode = unsafeCoerce
myDecode :: ByteString -> Text
myDecode = unsafeCoerce
This is a great idea if you enjoy segfault ...
This question implicates a sprawling range of issues. I'm going to interpret it as "In Haskell, how should I convert between Unicode and other character encodings?"
In Haskell, the recommended way to convert to and from Unicode is with the functions in text-icu, which provides some basic functions:
fromUnicode :: Converter -> Text -> ByteString
toUnicode :: Converter -> ByteString -> Text
text-icu is a binding to the International Components for Unicode libraries, which does the heavy lifting for, among other things, encoding and decoding to non-Unicode character sets. Its website gives documentation on conversion in general and some specific information on how its converter implementations operate. Note that different character sets require somewhat different coverter implementations.
ICU can also attempt to automatically detect the character set of an input. "This is, at best, an imprecise operation using statistics and heuristics." No other implementation could "fix" that characteristic. The Haskell bindings do not expose that functionality as I write; see #8.
I don't know of any character set conversion procedures written in native Haskell. As the ICU documentation indicates, there is a lot of complexity; after all, this is a rich area of international computing history.
Performance
As the ICU FAQ laconically notes, "Most of the time, the memory throughput of the hard drive and RAM is the main performance constraint." Although that comment is not specifically about conversions, I'd expect it to be broadly the case here as well. Is your experience otherwise?
unsafeCoerce is not appropriate here.
Is there any difference in how Data.Text and Data.Vector.Unboxed Char work internally? Why would I choose one over the other?
I always thought it was cool that Haskell defines String as [Char]. Is there a reason that something analagous wasn't done for Text and Vector Char?
There certainly would be an advantage to making them the same.... Text-y and Vector-y tools could be written to be used in both camps. Imagine Ropes of Ints, or Regexes on strings of poker cards.
Of course, I understand that there were probably historical reasons and I understand that most current libraries use Data.Text, not Vector Char, so there are many practical reasons to favor one over the other. But I am more interested in learning about the abstract qualities, not the current state that we happen to be in.... If the whole thing were rewritten tomorrow, would it be better to unify the two?
Edit, with more info-
To put stuff into perspective-
According to this page, http://www.haskell.org/haskellwiki/GHC/Memory_Footprint, GHC uses 16 bytes for each Char in your program!
Data.Text is not O(1) index'able, it is O(n).
Ropes (binary trees wrapped around text) can also hold strings.... They have better complexity for index/insert/delete, although depending on the number of nodes and balance of the tree, index could be close to that of Text.
This is my takeaway from this-
Text and Vector Char are different internally....
Use String if you don't care about performance.
If performance is important, default to using Text.
If fast indexing of chars is necessary, and you don't mind a lot of memory overhead (up to 16x), use Vector Char.
If you want to insert/delete a lot of data, use Ropes.
It's a fairly bad idea to think of Text as being a list of characters. Text is designed to be thought of as an opaque, user-readable blob of Unicode text. Character boundaries might be defined based on encoding, locale, language, time of month, phase of the moon, coin flips performed by a blinded participant, and migratory patterns of Venezuela's national bird whatever it may be. The same story happens with sorting, up-casing, reversing, etc.
Which is a long way of saying that Text is an abstract type representing human language and goes far out of its way to not behave just the same way as its implementation, be it a ByteString, a Vector UTF16CodePoint, or something totally unique (which is the case).
To clarify this distinction take note that there's no guarantee that unpack . pack witnesses an isomorphism, that the preferred ways of converting from Text to ByteString are in Data.Text.Encoding and are partial, and that there's a whole sophisticated plug-in module text-icu littered with complex ways of handling human language strings.
You absolutely should use Text if you're dealing with a human language string. You should also be really careful to treat it with care since human language strings are not easily amenable to computer processing. If your string is better thought of as a machine string, you probably should use ByteString.
The pedagogical advantages of type String = [Char] are high, but the practical advantages are quite low.
To add to what J. Abrahamson said, it's also worth making the distinction between iterating over runes (roughly character by character, but really could be ideograms too) as opposed to unitary logical unicode code points. Sometimes you need to know if you're looking at a code point that has been "decorated" by a previous code point.
In the case of the latter, you then have to make the distinction between code points that stand alone (such as letters, ideograms) and those that modify the text that follows (right-to-left code point, diacritics, etc).
Well implemented unicode libraries will typically abstract these details away and let you process the text in a more or less character-by-character fashion but you have to drop certain assumptions that come from thinking in terms of ASCII.
A byte is not a character. A logical unit of text isn't necessarily a "character". Not every code point stands alone, some decorate/annotate the following code point or even the rest of the byte stream until invalidated (right-to-left).
Unicode is hard. There is no one true encoding that will eliminate the difficulty of encapsulating the variety inherent in human language. Data.Text does a respectable job of it though.
To summarize:
The methods of processing are:
byte-by-byte - totally invalid for unicode, only applicable to latin-1/ASCII
code point by code point - works for processing unicode, but is lower-level than people realize
logical rune-by-rune - what you actually want
The types are:
String (aka [Char]) - has a limited scope. Best used for teaching Haskell or for legacy use-cases.
Text - the preferred way to handle "human" text.
Bytestring - for byte streams, raw data, binary etc.
The Unicode Normalization FAQ includes the following paragraph:
Programs should always compare canonical-equivalent Unicode strings as equal ... The Unicode Standard provides well-defined normalization forms that can be used for this: NFC and NFD.
and continues...
The choice of which to use depends on the particular program or system. NFC is the best form for general text, since it is more compatible with strings converted from legacy encodings. ... NFD and NFKD are most useful for internal processing.
My questions are:
What makes NFC best for "general text." What defines "internal processing" and why is it best left to NFD? And finally, never minding what is "best," are the two forms interchangable as long as two strings are compared using the same normalization form?
The FAQ is somewhat misleading, starting from its use of “should” followed by the inconsistent use of “requirement” about the same thing. The Unicode Standard itself (cited in the FAQ) is more accurate. Basically, you should not expect programs to treat canonically equivalent strings as different, but neither should you expect all programs to treat them as identical.
In practice, it really depends on what your software needs to do. In most situations, you don’t need to normalize at all, and normalization may destroy essential information in the data.
For example, U+0387 GREEK ANO TELEIA (·) is defined as canonical equivalent to U+00B7 MIDDLE DOT (·). This was a mistake, as the characters are really distinct and should be rendered differently and treated differently in processing. But it’s too late to change that, since this part of Unicode has been carved into stone. Consequently, if you convert data to NFC or otherwise discard differences between canonically equivalent strings, you risk getting wrong characters.
There are risks that you take by not normalizing. For example, the letter “ä” can appear as a single Unicode character U+00E4 LATIN SMALL LETTER A WITH DIAERESIS or as two Unicode characters U+0061 LATIN SMALL LETTER A U+0308 COMBINING DIAERESIS. It will mostly be the former, i.e. the precomposed form, but if it is the latter and your code tests for data containing “ä”, using the precomposed form only, then it will not detect the latter. But in many cases, you don’t do such things but simply store the data, concatenate strings, print them, etc. Then there is a risk that the two representations result in somewhat different renderings.
It also matters whether your software passes character data to other software somehow. The recipient might expect, due to naive implicit assumptions or consciously and in a documented manner, that its input is normalized.
NFC is the general common sense form that you should use, ä is 1 code point there and that makes sense.
NFD is good for certain internal processing - if you want to make accent-insensitive searches or sorting, having your string in NFD makes it much easier and faster. Another usage is making more robust slug titles. These are just the most obvious ones, I am sure there are plenty of more uses.
If two strings x and y are canonical equivalents, then
toNFC(x) = toNFC(y)
toNFD(x) = toNFD(y)
Is that what you meant?
I'm currently teaching myself Haskell, and I'm wondering what the best practices are when working with strings in Haskell.
The default string implementation in Haskell is a list of Char. This is inefficient for file input-output, according to Real World Haskell, since each character is separately allocated (I assume that this means that a String is basically a linked list in Haskell, but I'm not sure.)
But if the default string implementation is inefficient for file i/o, is it also inefficient for working with Strings in memory? Why or why not? C uses an array of char to represent a String, and I assumed that this would be the default way of doing things in most languages.
As I see it, the list implementation of String will take up more memory, since each character will require overhead, and also more time to iterate over, because a pointer dereferencing will be required to get to the next char. But I've liked playing with Haskell so far, so I want to believe that the default implementation is efficient.
Apart from String/ByteString there is now the Text library which combines the best of both worlds—it works with Unicode while being ByteString-based internally, so you get fast, correct strings.
Best practices for working with strings performantly in Haskell are basically: Use Data.ByteString/Data.ByteString.Lazy.
http://hackage.haskell.org/packages/archive/bytestring/latest/doc/html/
As far as the efficiency of the default string implementation goes in Haskell, it's not. Each Char represents a Unicode codepoint which means it needs at least 21bits per Char.
Since a String is just [Char], that is a linked list of Char, it means Strings have poor locality of reference, and again means that Strings are fairly large in memory, at a minimum it's N * (21bits + Mbits) where N is the length of the string and M is the size of a pointer (32, 64, what have you) and unlike many other places where Haskell uses lists where other languages might use different structures (I'm thinking specifically of control flow here), Strings are much less likely to be able to be optimized to loops, etc. by the compiler.
And while a Char corresponds to a codepoint, the Haskell 98 report doesn't specify anything about the encoding used when doing file IO, not even a default much less a way to change it. In practice GHC provides an extensions to do e.g. binary IO, but you're going off the reservation at that point anyway.
Even with operations like prepending to front of the string it's unlikely that a String will beat a ByteString in practice.
The answer is a bit more complex than just "use lazy bytestrings".
Byte strings only store 8 bits per value, whereas String holds real Unicode characters. So if you want to work with Unicode then you have to convert to and from UTF-8 or UTF-16 all the time, which is more expensive than just using strings. Don't make the mistake of assuming that your program will only need ASCII. Unless its just throwaway code then one day someone will need to put in a Euro symbol (U+20AC) or accented characters, and your nice fast bytestring implementation will be irretrievably broken.
Byte strings make some things, like prepending to the start of a string, more expensive.
That said, if you need performance and you can represent your data purely in bytestrings, then do so.
The basic answer given, use ByteString, is correct. That said, all of the three answers before mine have inaccuracies.
Regarding UTF-8: whether this will be an issue or not depends entirely on what sort of processing you do with your strings. If you're simply treating them as single chunks of data (which includes operations such as concatenation, though not splitting), or doing certain limited byte-based operations (e.g., finding the length of the string in bytes, rather than the length in characters), you won't have any issues. If you are using I18N, there are enough other issues that simply using String rather than ByteString will start to fix only a very few of the problems you'll encounter.
Prepending single bytes to the front of a ByteString is probably more expensive than doing the same for a String. However, if you're doing a lot of this, it's probably possible to find ways of dealing with your particular problem that are cheaper.
But the end result would be, for the poster of the original question: yes, Strings are inefficient in Haskell, though rather handy. If you're worried about efficiency, use ByteStrings, and view them as either arrays of Char8 or Word8, depending on your purpose (ASCII/ISO-8859-1 vs Unicode of some sort, or just arbitrary binary data). Generally, use Lazy ByteStrings (where prepending to the start of a string is actually a very fast operation) unless you know why you want non-lazy ones (which is usually wrapped up in an appreciation of the performance aspects of lazy evaluation).
For what it's worth, I am building an automated trading system entirely in Haskell, and one of the things we need to do is very quickly parse a market data feed we receive over a network connection. I can handle reading and parsing 300 messages per second with a negligable amount of CPU; as far as handling this data goes, GHC-compiled Haskell performs close enough to C that it's nowhere near entering my list of notable issues.