Re: Is there a reader setting support BigDecimal by default?
On Tue, Oct 18, 2011 at 3:47 PM, Tassilo Horn tass...@member.fsf.org wrote: Scott Hickey jscotthic...@gmail.com writes: And usually, you should refrain from using floating points at all, no matter if BigDecimal or Double. I thought BigDecimal with was not a floating point in the traditional sense (ie., subject to all of the usual rounding horror, unless you ask it to be)? That is, you can do decimal calculations exactly using it. -- [what were the skies like when you were young?] -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
On Thu, Oct 20, 2011 at 11:17, Joel Gluth joel.gl...@gmail.com wrote: On Tue, Oct 18, 2011 at 3:47 PM, Tassilo Horn tass...@member.fsf.org wrote: Scott Hickey jscotthic...@gmail.com writes: And usually, you should refrain from using floating points at all, no matter if BigDecimal or Double. I thought BigDecimal with was not a floating point in the traditional sense (ie., subject to all of the usual rounding horror, unless you ask it to be)? That is, you can do decimal calculations exactly using it. The term floating point is used to distingiush from fixed-point schemes, such as deciding that We're going to store money as a 64 bit integer representing the number of 100ths of a cent. Effectively, fixing the decimal point thus: $1000.. BigDecimal is an arbitrary precision integer and a scale factor (think of the scale factor as multiplying or dividing by a power of ten.) So, in that sense the (decimal) point floats. Unlike floating point in the IEEE754 sense, BigDecimal is decimal, not binary, (so e.g. 0.1 has a terminating representation) and arbitrary precision, so you can avoid rounding effects unless you're dividing where you might have to round to truncate a non-terminating representation. // Ben -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
Joel Gluth joel.gl...@gmail.com writes: And usually, you should refrain from using floating points at all, no matter if BigDecimal or Double. I thought BigDecimal with was not a floating point in the traditional sense (ie., subject to all of the usual rounding horror, unless you ask it to be)? That is, you can do decimal calculations exactly using it. You are right. What I've meant that you don't want to use double or float because it's not precise, and you don't want to use BigDecimal because it's slower and, even more important, it'll throw an exception if a given number cannot be represented as decimal number. user= (/ 1M 3) ArithmeticException Non-terminating decimal expansion; no exact representable decimal result. java.math.BigDecimal.divide (BigDecimal.java:1616) So the general suggestion is to stick to integers and rational numbers as long as possible. Bye, Tassilo -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
Thank for the replies and I appreciate the suggestions, however they some of the rationale behind them doesn't match well my experience. First, BigDecimal is plenty fast the large business systems I've worked on. Actually, it has been plenty fast for every large business system I've worked on. This includes a group insurance rate engine that generated hundreds of thousands of calculations in each web request. Second, storing values with implied decimal points is a nightmare. In the insurance application for example, some of the rate tables have precision to three decimal places, others to five. Picking some arbitrary value imposes code complexity throughout the whole in app in a very nasty way. Of course, what happens when you need to change that value two years from now? Unless there's a really good performance issue for a given application, I would never pick implied decimal representation over BigDecimals. Finally, for the business applications I've worked with, I haven't had to worry about representing all rationals, just base 10 numbers. In my experience, using BigDecimal by default for any number with a decimal point has worked out very well for balancing the needs of accuracy, speed and code complexity. It appears that the answer to the original question is no, there is no way to configure the reader to default numbers with a decimal point to be BigDecimal instead of Double. Scott Hickey -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
Scott: no, there is no way to configure the reader to default numbers with a decimal point to be BigDecimal instead of Double Correct. But you can modify the Reader: it's just Java code. -S -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
It appears that the answer to the original question is no, there is no way to configure the reader to default numbers with a decimal point to be BigDecimal instead of Double. Scott Hickey Reading a double implies that somebody upstream of you was using doubles, which violates the guarantees you want from BigDecimals. Why is the upstream provider using doubles? Stu -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
On Thu, Oct 20, 2011 at 22:50, Stuart Halloway stuart.hallo...@gmail.com wrote: It appears that the answer to the original question is no, there is no way to configure the reader to default numbers with a decimal point to be BigDecimal instead of Double. Scott Hickey Reading a double implies that somebody upstream of you was using doubles, which violates the guarantees you want from BigDecimals. Why is the upstream provider using doubles? I don't follow. The OP has text, which Clojure is reading as doubles. This only implies that upstream (which need not have been written in Clojure) is producing numbers matching #[-]?[1-9][0-9]*[.][0-9]*|0, because LispReader interprets that as Double. Whatever internal representation this text was produced form may or may not have been (binary) floating point initially. It doesn't seem reasonable to assume that the OP's 'business' applications I've built over the last 25 years could have known that Clojure would come along later and expect to find M on the end of every decimal number. // Ben // Ben -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
It appears that the answer to the original question is no, there is no way to configure the reader to default numbers with a decimal point to be BigDecimal instead of Double. Scott Hickey Reading a double implies that somebody upstream of you was using doubles, which violates the guarantees you want from BigDecimals. Why is the upstream provider using doubles? I don't follow. The OP has text, which Clojure is reading as doubles. This only implies that upstream (which need not have been written in Clojure) is producing numbers matching #[-]?[1-9][0-9]*[.][0-9]*|0, because LispReader interprets that as Double. Whatever internal representation this text was produced form may or may not have been (binary) floating point initially. It doesn't seem reasonable to assume that the OP's 'business' applications I've built over the last 25 years could have known that Clojure would come along later and expect to find M on the end of every decimal number. // Ben Hmm, highly apropos discussion since Rich's talk on simplicity posted today. The reader does one thing: read Clojure data. Sometimes you need something else: read data written in another format. You might accomplish that by (1) making the reader able to do two things, adding a flag to deal with a specific format, or (2) by using a different reader for that job. (2) is hands-down the right answer. Stu -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
On Thu, Oct 20, 2011 at 23:16, Stuart Halloway stuart.hallo...@gmail.com wrote: It appears that the answer to the original question is no, there is no way to configure the reader to default numbers with a decimal point to be BigDecimal instead of Double. Scott Hickey Reading a double implies that somebody upstream of you was using doubles, which violates the guarantees you want from BigDecimals. Why is the upstream provider using doubles? I don't follow. The OP has text, which Clojure is reading as doubles. This only implies that upstream (which need not have been written in Clojure) is producing numbers matching #[-]?[1-9][0-9]*[.][0-9]*|0, because LispReader interprets that as Double. Whatever internal representation this text was produced form may or may not have been (binary) floating point initially. It doesn't seem reasonable to assume that the OP's 'business' applications I've built over the last 25 years could have known that Clojure would come along later and expect to find M on the end of every decimal number. // Ben Hmm, highly apropos discussion since Rich's talk on simplicity posted today. The reader does one thing: read Clojure data. Sometimes you need something else: read data written in another format. You might accomplish that by (1) making the reader able to do two things, adding a flag to deal with a specific format, or (2) by using a different reader for that job. (2) is hands-down the right answer. No argument there. The Clojure reader's job is to read Clojure. If your have some data that happens to be syntactically compatible, fine, but if not the Clojure reader isn't the right tool. One obvious approach would be a to use a regular expression to identify the numbers in the text to be parsed and #(BigDecimal. ^String %) to do the actual conversion. // Ben -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
I'm looking to avoid qualifying every number that has a decimal point with an M. For the business applications I've built over the last 25 years (credit card processing, healthcare claims, loans,etc.), there's never been a situation where inexact numbers were appropriate. That's why I was hoping for a global default for reading decimal numbers as BigDecimal instead of IEEE floats or doubles. Scott Hickey -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
Scott Hickey jscotthic...@gmail.com writes: Hi Scott, I'm looking to avoid qualifying every number that has a decimal point with an M. For the business applications I've built over the last 25 years (credit card processing, healthcare claims, loans,etc.), there's never been a situation where inexact numbers were appropriate. Nobody wants inexact numbers, but one has to trade speed with exactness. And usually, you should refrain from using floating points at all, no matter if BigDecimal or Double. Clojure has rationals, so unless you have to use java.lang.Math's functions or PI and E constants, you can possibly do all calculations with rationals, and when you really need a float, use (bigdec my-rational) to coerce it to a BigDecimal. That's why I was hoping for a global default for reading decimal numbers as BigDecimal instead of IEEE floats or doubles. I've just checked, and that's hard-coded in clojure.lang.LispReader. Bye, Tassilo -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Is there a reader setting support BigDecimal by default?
In some versions of Scheme or Lisp, there is a flag that you can set so that the reader will create exact numbers by default (BigDecimal) instead of inexact doubles. Is there a way to do this in Clojure? Scott Hickey -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
Hi, you are looking for 1.3M? Sincerely Meikel -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
I've been using 1.2 but I would be OK with a solution in any version. Scott Hickey -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
Re: Is there a reader setting support BigDecimal by default?
Hi, I meant this: Clojure 1.2.0 user= (type 1.0M) java.math.BigDecimal Sincerely Meikel -- You received this message because you are subscribed to the Google Groups Clojure group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en
