I am working through the first edition of this book and while I enjoy it, some of the examples given seem out-dated. I would give up and find another book to learn from, but I am really interested in what the author is talking about and want to make the examples work for myself, so I am trying to update them as I go along.
The following code is a map/reduce approach to analyzing text that depends on clojure.contrib. I have tried changing the .split function to re-seq with #"\w+", used line-seq instead of read-lines, and changed the .toLowerCase to string/lower-case. I tried to follow my problems to the source code and read the docs thoroughly to learn that the read-lines function closes after you consume the entire sequence and that line-seq returns a lazy sequence of strings, implementing java.io.BufferedReader. The most helpful thing for my problem was post about how to read files after clojure 1.3. Even still, I can't get it to work.
So here's my question: What dependencies and/or functions do I need to change in the following code to make it contemporary, reliable, idiomatic Clojure?
First namespace:
(ns chapter-data.word-count-1
(:use clojure.contrib.io
clojure.contrib.seq-utils))
(defn parse-line [line]
(let [tokens (.split (.toLowerCase line) " ")]
(map #(vector % 1) tokens)))
(defn combine [mapped]
(->> (apply concat mapped)
(group-by first)
(map (fn [[k v]]
{k (map second v)}))
(apply merge-with conj)))
(defn map-reduce [mapper reducer args-seq]
(->> (map mapper args-seq)
(combine)
(reducer)))
(defn sum [[k v]]
{k (apply + v)})
(defn reduce-parsed-lines [collected-values]
(apply merge (map sum collected-values)))
(defn word-frequency [filename]
(map-reduce parse-line reduce-parsed-lines (read-lines filename)))
Second namespace:
(ns chapter-data.average-line-length
(:use rabbit-x.data-anal
clojure.contrib.io))
(def IGNORE "_")
(defn parse-line [line]
(let [tokens (.split (.toLowerCase line) " ")]
[[IGNORE (count tokens)]]))
(defn average [numbers]
(/ (apply + numbers)
(count numbers)))
(defn reducer [combined]
(average (val (first combined))))
(defn average-line-length [filename]
(map-reduce parse-line reducer (read-lines filename)))
But when I compile and run it in light table I get a bevy of errors:
1) In the word-count-1 namespace I get this when I try to reload the ns function after editing:
java.lang.IllegalStateException: spit already refers to: #'clojure.contrib.io/spit in namespace: chapter-data.word-count-1
2) In the average-line-length namespace I get similar name collision errors under the same circumstances:
clojure.lang.Compiler$CompilerException: java.lang.IllegalStateException: parse-line already refers to: #'chapter-data.word-count-1/parse-line in namespace: chapter-data.average-line-length, compiling:(/Users/.../average-line-length.clj:7:1)
3) Oddly, when I quit and restart light table, copy and paste the code directly into the files (replacing what's there) and call instances of their top level functions the word-count-1 namespace runs fine, giving me the number of occurrences of certain words in the test.txt file but the average-line-length name-space gives me this:
"Warning: *default-encoding* not declared dynamic and thus is not dynamically rebindable, but its name suggests otherwise. Please either indicate ^:dynamic *default-encoding* or change the name. (clojure/contrib/io.clj:73)...
4) At this point when I call the word-frequency functions of the first namespace it returns nil instead of the number of word occurrences and when I call the average-line-length function of the second namespace it returns
java.lang.NullPointerException: null
core.clj:1502 clojure.core/val
As far as I can tell, clojure.contrib.io and clojure.contrib.seq-utils are no longer updated, and in fact they may be conflicting with clojure.core functions like spit. I would recommend taking out those dependencies and seeing if you can do this using only core functions. spit should just work -- the error that you're getting is caused by useing clojure.contrib.io, which contains its own spit function, which looks to be roughly equivalent; perhaps the current version in clojure.core is a "new and improved" version of clojure.contrib.io/spit.
Your problem with the parse-line function looks to be caused by the fact that you've defined two functions with the same name, in two different namespaces. The namespaces don't depend on one another, but you can still run into a conflict if you load both namespaces in a REPL. If you only need to use one at a time, try using one of them, and then when you want to use the other one, make sure you do a (remove-ns name-of-first-ns) first to free up the vars so there is no conflict. Alternatively, you could make parse-line a private function in each namespace, by changing (defn parse-line ... to (defn- parse-line ....
EDIT: If you still need any functions that were in clojure.contrib.io or clojure.contrib.seq-utils that aren't available in core or elsewhere, you can always copy the source over into your namespace. See clojure.contrib.io and clojure.contrib.seq-utils on github.
Related
for an assignment I need to create a map from a text file in clojure, which I am new to. I'm specifically using a hash-map...but it's possible I should be using another type of map. I'm hoping someone here can answer that for me. I did try changing my hash-map to sorted-map but it gave me the same problem.
The first character in every line in the file is the key and the whole line is the value. The key is a number from 0-9999. There are 10,000 lines and each number after the first number in a line is a random number between 0 and 9999.
I've created the hashmap successfully I think. At least, its not giving me an error when I just run that code. However when I try to iterate through it, printing every value for keys 0-9999 it gives me a stack overflow error right at the middle of line 2764(in the text file). I'm hoping someone can tell me why it's doing this and a better way to do it?
Here's my code:
(ns clojure-project-441.core
(:gen-class))
(defn -main
[& args]
(def pages(def hash-map (file)))
(iter 0)
)
(-main)
(defn file []
(with-open [rdr (clojure.java.io/reader "pages.txt")]
(reduce conj [] (line-seq rdr))))
(defn iter [n]
(doseq [keyval (pages n)] (print keyval))
(if (< n 10000)
(iter (inc n))
)
)
here's a screenshot of my output
If it's relevant at all I'm using repl.it as my IDE.
Here are some screenshots of the text file, for clarity.
beginning of text file
where the error is being thrown
Thanks.
I think the specific problem that causes the exception to be thrown is caused because iter calls itself recursively too many times before hitting the 10,000 line limit.
There some issues in your code that are very common to all people learning Clojure; I'll try to explain:
def is used to define top-level names. They correspond with the concept of constants in the global scope on other programming languages. Think of using def in the same way you would use defn to define functions. In your code, you probably want to use let to give names to intermediate results, like:
(let [uno 1
dos 2]
(+ uno dos)) ;; returns 3
You are using the name hash-map to bind it to some result, but that will get in the way if you want to use the function hash-map that is used to create maps. Try renaming it to my-map or similar.
To call a function recursively without blowing the stack you'll need to use recur for reasons that are a bit long to explain. See the factorial example here: https://clojuredocs.org/clojure.core/recur
My advice would be to think of this assignment as a pipeline composed of the following small functions:
A function that reads the lines from the file (you already have this)
A function that, given a line, returns a pair: the first element of the pair is the first number of the line, the second element is the whole line (the input parameter) OR
A function that reads the first number of the line
To build the map, you have a few options; two off the top of my mind:
Use a loop construct and, for each line, "update" the hash-map to include a new key-value pair (the key is the first number, the value is the whole line), then return the whole hash-map you've built
Use a reduce operation: you create a collection of key-value pairs, then tell reduce to merge, one step at a time, into the original hash-map. The result is the hash-map you want
I think the key is to get familiar with the functions that you can use and build small functions that you can test in isolation and try to group them conveniently to solve your problem. Try to get familiar with functions like hash-map, assoc, let, loop and recur. There's a great documentation site at https://clojuredocs.org/ that also includes examples that will help you understand each function.
I am trying to 'purify' some of my Clojure functions. I would like to get to a point where all my side-effecting code is explicitly declared in one function. It's easy to get some data at the start and to write it to a db at the end and just have a pure function transforming in between. However, the usual situation is that the transformation function requires another DB read somewhere in the middle of the logic:
(defn transform-users
[users]
(let [ids (map :id users)
profiles (db/read :profiles ids)]
(profiles->something profiles)))
(->> (db/read :users)
(transform-users)
(db/write :something)
Obviously this is a very simple example but the point is, how do I get the side-effecting db/read function out of there, how can I make transform-users pure (and as a benefit, easily testable)?
One thing you could do here would be a dependency-injection-like approach of supplying the (potentially) side-effectful function as an optional parameter, e.g.:
(defn transform-users
[users & {:keys [ids->profiles]
:or {ids->profiles #(db/read :profiles %)}]
(let [ids (map :id users)
profiles (ids->profiles ids)]
(profiles->something profiles)))
This should be easily testable since you can mock the injected functions without a lot of effort. And as a bonus, by supplying the default value, you're documenting what you're expecting and making the function convenient to call.
Why couple the reading of the profiles with transforming profiles?
(->> (db/read :users)
(map :id)
(db/read :profiles)
(profile->something)
(db/write :something)
(This also exposes the fact that you are doing two round trips to the db. Where is db/read :user-profiles ?)
(->> (db/read :user-profiles)
(profile->something)
(db/write :something)
or perhaps:
(->> (read-profiles-from-users)
(profile->something)
(db/write :something)
I am new to Clojure and am just trying to build some sample apps to get used to the syntax. I noticed the following order dependency behaviour.
I created a project called timex to calculate the time in weeks between two dates. I am using the clj-time functions for the date difference calculations.
If my core.clj looks as follows:
(ns timex.core
(:gen-class))
(defn -main
"Calculate weeks between dates."
[& args]
dp
)
(require '[clj-time.core :as t])
(def d2 (t/date-time 1989 01 07))
(def dw (t/in-weeks (t/interval d2 (t/now))))
(def dp (str "The number of weeks between Jan 7, 1989 and now is " dw "!"))
**
If I run lein repl I get the following error:
#CompilerException java.lang.RuntimeException: Unable to resolve symbol: dp in this context, compiling:(timex/core.clj:4:1)
But if I re-order the lines in the file and put the def's and the require statement before the main as such
(ns timex.core
(:gen-class))
(require '[clj-time.core :as t])
(def d2 (t/date-time 1989 01 07))
(def dw (t/in-weeks (t/interval d2 (t/now))))
(def dp (str "The number of weeks between Jan 7, 1989 and now is " dw "!"))
(defn -main
"Calculate weeks between dates."
[& args]
dp
)
**
Then when I run lein repl and the invoke the (-main) function, I get:
timex.core=> (-main)
"The number of weeks between Jan 7, 1989 and now is 1341!"
Is this apparent order-dependency normal or am I doing something wrong? If the latter, then I would appreciate any advice or documentation I should review. Thanks.
The unit of compilation in Clojure is the s-expression as opposed to a whole file like many other languages. When you load a file that file file is evaluated from top to bottom in a single pass. vars (what is created by calls to def) must be created above where they are used, or you can use declare if you need mutual recursion though this is rare.
It's worth spending some time getting used to the difference between compiling a file in the traditional sense, and loading a file in the lisp sense because it is fundamental to the macro system and many other aspects of the language. When you require a file from a namespace, or call load-file from the repl the clojure compiler is invoked and repeatedly reads, macro-expands, then evaluates each from in the file starting at the top. The first line tels it what namespace to define things in which is why the ns expression comes first. Then further forms define things in that namespace. If you then load the file again it does nothing more than read the file from the top all over again. It does not clear the namespace or any other magic, so if you evaluate the file, then change the order and evaluate it again it can continue to work because the required functions are already defined in the namespace (as stored in memeory) when the second pass is run. When getting used to this it helps to run lein check often to make sure things are well ordered.
PS: the vast majority of the time the call to require goes in the ns form at the top of the file.
Having a collection of files/directories I'd like to create a collection of all leaves files.
I'd like the resulting collection extend the clojure.core.protocols/CollReduce protocol.
Do I need to extend the protocol or are there helper functions for that?
In other words. Do reducers only help in parallel reduction or can I use them to effectively generate reducible collections in parallel as well?
To illustrate the problem let me show the implementation which could work, granted that the file hierarchy is no deeper than two levels (e.g our collection can contain files and directories, but the directories can contain only files)
(ns user
[import [java.io File]])
(defn expand [reduction-function]
(fn [result input]
(if (.isFile input)
(reduction-function result input)
; if not a file we assume it's a directory
(reduce reduction-function result (.listFiles input)))))
(defn process [xfn c]
(lazy-seq (when-let [s (seq c)]
(concat ((xfn #(concat %1 (list %2))) '() (first s))
(process xfn (rest s))))))
(def f (File. "C:\\WORK"))
(process expand [f]) ; => produces list of files
Now, it would be nice to have the expand defined in recursion-like style (or rather as a cascade of transformations), so it works for all levels, but executed in parallel fashion. As with reducers we can define early termination in reduction, I'd like to have ability to define generation (collection expansion) which stops when certain criteria is met (reaching file in directory hierarchy tree, as in file example)
So I am using congomongo (the fetch function near the end) to pull some documents from a mongodb collection. I want to pass options through to the fetch call, so i can do something like (posts :limit 1) and have {:limit 1} get passed through to fetch. I am doing hand rolled "memoization" with #posts, because i want to be able to reset the cache, which to my understanding cant be done with clojure.core/memoize.
Now, the problem I see here is that (fetch :posts options) call is non trivial, and I would really rather not hammer my datastore if dosync has to retry the transaction. I am a total clojure/fp noob though, and I am not sure how to get around that problem. Also, since I am a noob, if I am doing anything else here that makes you cringe, I would love to find out how to write this properly.
(def posts (ref nil))
(defn reset-posts [] (dosync alter posts nil))
(defn fetch-posts [& options]
(let [options (apply array-map options)]
(or #posts
(dosync alter posts (fetch :posts options)))))
I'm not convinced that your transaction blocks ((dosync alter...) do what you think!
user=> (def posts (ref nil))
#'user/posts
user=> (dosync (ref-set posts [1 2 3 4 5]))
[1 2 3 4 5]
user=> #posts
[1 2 3 4 5]
user=> (dosync alter posts nil)
nil
user=> #posts
[1 2 3 4 5]
In reset-posts, you probably want (dosync (ref-set posts nil)), and in fetch-posts, the syntax fix would be (dosync (ref-set posts (fetch :posts options))).
However, there's a race condition in fetch-posts, a check-then-act. Might not be that big of a deal; not sure who uses fetch-posts, but moving the or #posts bit inside the transaction would avoid a situation where 2 concurrent transactions both end up committing the alter.
With regard to retries of fetch-posts, yeah, that could happen, though your cache solution avoids most of them. I'm not sure there's a way around it without locking, though. Usually with I/O stuff in transactions you'd farm it out to an agent, but the transaction's success depends on the return value from fetch, so it's not clear to me how that'd work.
So you're introducing the ref because you want to be able to not blow up memory when time passes, 'cause just using memoize around fetch-posts may lead to this, sooner or later, right ?
Maybe you could try an alternate approach : let fetch-posts be "pure", memoize-free. With this scenario, someone can call fetch-posts blindlessly, without having to fear OutOfMemoryExceptions.
Indeed, maybe for some usecases, it may be sufficient to "cache the value" in a local of the calling code.
But the story does not end here, or I would not have taken the time to answer :-) : you can pretty easily have your "localized in time" memoize by rebinding fetch-posts using clojure.core/binding : from then, all the code in the same thread in the call stack will benefit from the bound memoized fetch-posts.
If you're using clojure 1.3 alpha, you'll need to declare the fetch-posts var as rebindable explicitly via the :dynamic metadata.
;; most simple definition
(defn ^:dynamic fetch-posts [& options]
(let [options (apply array-map options)]
(fetch :posts options)))
;; a la carte caching by the calling code (lexically scoped)
(let [posts (apply fetch-posts options)] ...)
;; a la carte caching by the calling code (dynamically scoped)
(binding [fetch-posts (memoize fetch-posts)] ...)
My last guess would be that you'd want to "memoize" in posts, in your original version, by indexing the posts by a key which would be the options seq, right ? Some maybe your code was not right ? (or you made the assumption that fetch-posts would always be called with the same args over and over ?)
Another idea. Use an agent to serialize write-access to posts and then ensure the call to fetch is only done when it is nil :
(def posts (agent nil))
(defn reset-posts [] (send posts (constantly nil)))
(defn fetch-posts [& options]
(let [options (apply array-map options)]
(send-off posts #(or % (fetch :posts options)))
(await-for (Long/MAX_VALUE) posts)
#posts))
Another approach which might be useful to move extensive computations outside the dosync is to use delay.
(defn fetch-posts
[& options]
#(dosync (or #posts (ref-set posts (delay (apply fetch :posts options))))))
Also note that your original code is not thread-safe since you access the ref outside the dosync and modify it based on this value afterwards in the dosync. But the value might have changed already between the deref and the dosync. Eg. by another thread calling fetch-posts in parallel.
Also the agent approach is questionable, because you cannot reliably read an agent. The value you get is consistent, but the access is not synchronised. Consider Laurent's example: between await-for and the deref another thread might already call reset-posts and you get nil instead of the post data. In this example this is probably a) far fetched and b) maybe a case one has to consider anyway, but there might be other use cases where this introduces a subtle race condition in more critical code.
tl;dr: Be careful what you do! Clojure is not magically thread-safe. Reason thoroughly about your solution and be aware of the implications.