On Jun 14, 2012, at 10:06 AM, Harold Erbin <[email protected]> wrote:
> Hi, > > I'm creating a new topic on the implementation of unit systems into > sympy to gather all discussions under an explicit name ([1] and [2]). > > First a summary of the code: > 1. First of all, dimensions are defined as Dict in order to let the > possibility to add new dimensions. They combines themselves as elements > of a group. It will then be possible to implement general computations > using uniquely dimensions (dimensional analysis, etc.). This is totally > coherent without unit system, even if it's possible to use dimension > with them to add new possibilities. > 2. Then units are defined. Their operations are consistent without the > choice of an unit system (see below), but then it's quite useless except > in some specific cases (like defining a system). > 3. System can be viewed as vector space (or group, which is more used in > literature).They are constructed by using "canonical" units (and so > dimension), which have to be defined before. But when the system is > created, it's possible to use other units as a basis, which is > equivalent to change the basis of the vector space. This is the simplest > way even if at some point we have to introduce a favored base. > 4. Finally quantities are just a number with an unit. > > There is a real liberty to define unit systems which will be useful, > instead of having hundred of units, and so one can really use only the > system and the units he needs (for example astrophysicists can only use > units with ua, Earth or Sun masses, the theoricists will use natural > system, etc.). > > The main gap in this approach is that unit and quantity are almost the > same things, since the factor of the unit can be moved to the quantity, > and vice versa. I'm thinking how to improve this, maybe by keeping in > the factor's unit only what is necessary to give 1 when expressed in > terms of base units, or when it's a special unit (like liter...). > > I decided to let the possibility to add units or an unit with a > quantity, because it's simpler when using interactive shell (in my > opinion, but I was hesitating a lot) : > m + m = 2*m > 2*m = 3*m > so the unit m is interpreted as the quantity 1*m. > > Now I'm again working on the problem we were speaking about with Aaron, > which is the way that Mul combine the units, resulting in expressions > like (cf the test file for more details) > 149598206033.591*kg**(1/3)*kg**-1**(1/3)*m**3**(1/3)*s**-2**(1/3)*s**2**(1/3) > > I did not solve yet the problem we were discussing in issues 1940 and > 1941, because I was hoping that subclassing AtomicExpr could avoid to > edit the Mul class (by seeing the current implementation of units). > > I have still a great list of things to implement, but the core is almost > done (I have still to implement the constants and use them to define > unit system). > At the end, what is the best way to list all the things to still > do/improve? (I will not be able to do all.) Should I put them in comment > inside the code, or open one issue on github and list all the ideas? (or > open one issue for each?). First I should note that issues should be opened in Google Code, not GitHub. GitHub is just for pull requests. Either of those options is ok. A third option is to start a page on the wiki. Each way has its own advantages. Comments on the code will tell anyone who stars editing the module what needs to be done, and it has the least chance of being "lost" if it is there. However if you choose toput it somewhere else, you can just add a link in the code. Issues have a good level of permanence and have a nice discussion format. Also anyone wanting to report a bug will hopefully find your issue first. Wiki pages are good if you are scoping out a design. You can get nice formatting, and other people can chime in if they want by just editing the page. Aaron Meurer > > The current code can be found here on github [3] (in the module > physics/units). Tests shows some examples of the computations available. > > Best, > > Harold > > [1] https://groups.google.com/forum/?fromgroups#!topic/sympy/Vrvwf5CYmWw > [2] https://groups.google.com/forum/?fromgroups#!topic/sympy/tpxqGzbCiMk > [3] https://github.com/melsophos/sympy/tree/units > > -- > You received this message because you are subscribed to the Google Groups > "sympy" group. > To post to this group, send email to [email protected]. > To unsubscribe from this group, send email to > [email protected]. > For more options, visit this group at > http://groups.google.com/group/sympy?hl=en. > -- You received this message because you are subscribed to the Google Groups "sympy" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/sympy?hl=en.
