code review 125540043: go.blog: constants

go.blog: constants

[gri, 2014-08-20 00:32:10]

LGTM but several comments. As an aside: I was thinking it would be nice if one could use //line comments to force error message for the blog examples show up with line numbers matching the example, as in: http://play.golang.org/p/pj5jYFj8Xw Unfortunately, the error message also shows the real location in []'s. Perhaps we could filter those out for playground/etc. compiler errors? https://codereview.appspot.com/125540043/diff/20001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:34: When designing Go, we decided to avoid this whole minefield by mandating that there is _no_ mixing of numeric types. s/whole minefield/minefield entirely/ (?) reads more smoothly (to my ears) https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:63: Such values, named or otherwise, are called _constants_ in Go. There's no mention of constant expressions, but they are creeping in later, as in 1 + 0i. Perhaps just add `2 + 3` as another example? https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:86: This is an _untyped_string_constant_, which is to say it is a constant value holding some text, but it does not have a fixed type yet. a constant value of text or: a constant text value (the value is the text, it's not holding it) https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:155: One way to think about untyped constants is that they live in a kind of ideal space of values, a space not tied to Go's type system. s/tied to/constrained by/ or even s/tied to/much less constrained by/ Constants after all are very much tied to Go's type system, but it's much less restrictive. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:171: to pass as an argument, and the concrete type stored of that argument is the default type of the constant. s/of that/for that/ (?) https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:226: Numeric constants live in an arbitrary-precision numeric space; they are just abstract numbers. s/abstract// (?) I'd say an abstract number is 'x' while '42' is a concrete number... https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:237: But `Huge` might be useful, and we can compute with it and use the result of that computation if the result we can compute with it: another place where you have a constant expression - a careful reader not familiar with Go might wonder where this computation happens - before you said there's not much that can be done with constants https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:237: But `Huge` might be useful, and we can compute with it and use the result of that computation if the result s/can compute with it/express arithmetic with it/ (?) https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:245: In a related way, floating-point constants may have very high precision, so that calculations involving them are more accurate. so that arithmetic involving them is more accurate. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:261: soft underflows, and `NaNs` to arise in constant expressions. s/to // (?) https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:346: Here is an informative little exercise missing : at end https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:361: but that is illegal because the abstract number -1 cannot be represented by an unsigned variable; `-1` is not in the range of unsigned values. s/abstract // (?) https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:366: Even though at run-time a value of -1 can be converted to an unsigned integer, the rules of constants forbid this kind of coercion at compile time. s/of constants/for constant conversions/ https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:382: The solution is to tell Go how many bits to use to represent the `0`, and invert that. I've read this section now 3x, and each time I'm doing a double-take. The first part of the first sentence doesn't make sense to me (even though I know what you want to say), and the 2nd half doesn't explain why this solves the problem. The rest builds on this. How about: We know that the largest unsigned integer value has all `1` bits set. The prior attempts (-1, ^0) don't work because the operations (negation, inversion) operate on untyped numbers, resulting in untyped negative numbers which don't fit into a uint. So we must avoid negative values. One way to achieve that is to start with uint(0), the unsigned integer 0. We are now in "unsigned integer land", and flipping the bits using `^uint(0)` will result in another unsigned integer value of all 1's. (or some such) As an aside: In these examples you use constant expressions throughout (incl. conversions). Perhaps making clear that such expressions remain constants may be warranted earlier, despite attempts to keep it as simple as possible. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:404: All these constants have numeric value 1 missing : at end ?

[adg, 2014-08-20 01:30:39]

https://codereview.appspot.com/125540043/diff/20001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:167: func Printf(a ...interface{}) (n int, err error) you use Printf here but this signature is actually Print or Println's. Either the above example code should be fmt.Printf("%v\n", "Hello") or fmt.Println("Hello") https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:191: For string constants, it's obvious that the implicit type is `string`. how about "For string constants, the only possible implicit type is string." https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:193: Integer constants default to `int`, floating-point constants `float64`, rune constants to `rune` (an alias for `int32`), imaginary constants to `complex128`, and so on. link "alias' to http://golang.org/ref/spec#Numeric_types ? https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:289: prints `complex128:` `(2+0i)`. this fragment doesn't follow from the previous sentence https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:302: They have more moving parts—many sizes, signed or unsigned, and more—but they play by the same rules. maybe another place to link the numbers section of the spec https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:336: .play -edit constants/int3.go /var/ show a large number assignment, too? https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:348: The number of bits available depends on the architecture, so we can't just write down a single value. this sentence should be part of the text after the MaxUint32 decl https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:355: Fans of 2's-complement arithmetic, which Go's integers are defined to use, know that the representation of `-1` has all its bits set to 1, so the bit pattern of `-1` is internally the same as that of the link to http://en.wikipedia.org/wiki/Two's_complement ? https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:393: Whatever the number of bits it takes to represent a `uint` in the current execution environment (on the playground, it's 32), link to //blog.golang.org/playground

[r, 2014-08-20 16:06:20]

https://codereview.appspot.com/125540043/diff/20001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:34: When designing Go, we decided to avoid this whole minefield by mandating that there is _no_ mixing of numeric types. On 2014/08/20 00:32:10, gri wrote: > s/whole minefield/minefield entirely/ (?) > > reads more smoothly (to my ears) Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:63: Such values, named or otherwise, are called _constants_ in Go. On 2014/08/20 00:32:10, gri wrote: > There's no mention of constant expressions, but they are creeping in later, as > in 1 + 0i. Perhaps just add `2 + 3` as another example? Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:86: This is an _untyped_string_constant_, which is to say it is a constant value holding some text, but it does not have a fixed type yet. On 2014/08/20 00:32:10, gri wrote: > a constant value of text > > or: > > a constant text value > > (the value is the text, it's not holding it) Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:155: One way to think about untyped constants is that they live in a kind of ideal space of values, a space not tied to Go's type system. On 2014/08/20 00:32:10, gri wrote: > s/tied to/constrained by/ > > or even > > s/tied to/much less constrained by/ > > Constants after all are very much tied to Go's type system, but it's much less > restrictive. Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:167: func Printf(a ...interface{}) (n int, err error) On 2014/08/20 01:30:39, adg wrote: > you use Printf here but this signature is actually Print or Println's. > Either the above example code should be > fmt.Printf("%v\n", "Hello") > or > fmt.Println("Hello") Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:171: to pass as an argument, and the concrete type stored of that argument is the default type of the constant. On 2014/08/20 00:32:10, gri wrote: > s/of that/for that/ (?) Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:191: For string constants, it's obvious that the implicit type is `string`. On 2014/08/20 01:30:39, adg wrote: > how about "For string constants, the only possible implicit type is string." Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:193: Integer constants default to `int`, floating-point constants `float64`, rune constants to `rune` (an alias for `int32`), imaginary constants to `complex128`, and so on. On 2014/08/20 01:30:38, adg wrote: > link "alias' to http://golang.org/ref/spec#Numeric_types ? Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:226: Numeric constants live in an arbitrary-precision numeric space; they are just abstract numbers. On 2014/08/20 00:32:10, gri wrote: > s/abstract// (?) > > I'd say an abstract number is 'x' while '42' is a concrete number... Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:237: But `Huge` might be useful, and we can compute with it and use the result of that computation if the result On 2014/08/20 00:32:10, gri wrote: > s/can compute with it/express arithmetic with it/ (?) Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:237: But `Huge` might be useful, and we can compute with it and use the result of that computation if the result On 2014/08/20 00:32:10, gri wrote: > we can compute with it: another place where you have a constant expression - a > careful reader not familiar with Go might wonder where this computation happens > - before you said there's not much that can be done with constants Acknowledged. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:245: In a related way, floating-point constants may have very high precision, so that calculations involving them are more accurate. On 2014/08/20 00:32:10, gri wrote: > so that arithmetic involving them is more accurate. Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:261: soft underflows, and `NaNs` to arise in constant expressions. On 2014/08/20 00:32:10, gri wrote: > s/to // (?) Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:289: prints `complex128:` `(2+0i)`. On 2014/08/20 01:30:38, adg wrote: > this fragment doesn't follow from the previous sentence Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:302: They have more moving parts—many sizes, signed or unsigned, and more—but they play by the same rules. On 2014/08/20 01:30:38, adg wrote: > maybe another place to link the numbers section of the spec Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:336: .play -edit constants/int3.go /var/ On 2014/08/20 01:30:39, adg wrote: > show a large number assignment, too? that was done immediately above. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:346: Here is an informative little exercise On 2014/08/20 00:32:10, gri wrote: > missing : at end Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:348: The number of bits available depends on the architecture, so we can't just write down a single value. On 2014/08/20 01:30:38, adg wrote: > this sentence should be part of the text after the MaxUint32 decl Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:355: Fans of 2's-complement arithmetic, which Go's integers are defined to use, know that the representation of `-1` has all its bits set to 1, so the bit pattern of `-1` is internally the same as that of the On 2014/08/20 01:30:39, adg wrote: > link to http://en.wikipedia.org/wiki/Two ? Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:361: but that is illegal because the abstract number -1 cannot be represented by an unsigned variable; `-1` is not in the range of unsigned values. On 2014/08/20 00:32:10, gri wrote: > s/abstract // (?) Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:366: Even though at run-time a value of -1 can be converted to an unsigned integer, the rules of constants forbid this kind of coercion at compile time. On 2014/08/20 00:32:10, gri wrote: > s/of constants/for constant conversions/ Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:382: The solution is to tell Go how many bits to use to represent the `0`, and invert that. i don't think your version explains it very well and has its own issues. ^0 is not a negative number, for example. i've taken another crack at it. see if you like it better. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:393: Whatever the number of bits it takes to represent a `uint` in the current execution environment (on the playground, it's 32), On 2014/08/20 01:30:39, adg wrote: > link to //blog.golang.org/playground Done. https://codereview.appspot.com/125540043/diff/20001/content/constants.article... content/constants.article:404: All these constants have numeric value 1 On 2014/08/20 00:32:10, gri wrote: > missing : at end ? Done.

[r, 2014-08-20 17:54:21]

Hello gri@golang.org, adg@golang.org (cc: golang-codereviews@googlegroups.com), I'd like you to review this change to https://code.google.com/p/go.blog

[gri, 2014-08-20 17:58:06]

LGTM https://codereview.appspot.com/125540043/diff/40001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/40001/content/constants.article... content/constants.article:381: But that fails too, for a similar reason: ^0 represents an infinite number of 1s, and we lose information if we assign that to any fixed-size integer: perhaps, to make that a bit more obvious: s/: ^0/: In the ideal space of numeric values, ^0/

[adg, 2014-08-21 03:24:16]

LGTM https://codereview.appspot.com/125540043/diff/80001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/80001/content/constants.article... content/constants.article:2: 19 Aug 2014 Shall we post it on the 25th? (Monday)

[r, 2014-08-21 04:27:32]

SGTM. You'll do the submitting and posting, right? On Wed, Aug 20, 2014 at 8:24 PM, <adg@golang.org> wrote: > LGTM > > > https://codereview.appspot.com/125540043/diff/80001/content/constants.article > File content/constants.article (right): > > https://codereview.appspot.com/125540043/diff/80001/content/constants.article... > content/constants.article:2: 19 Aug 2014 > Shall we post it on the 25th? (Monday) > > https://codereview.appspot.com/125540043/ 

[rsc, 2014-08-22 00:39:36]

LGTM https://codereview.appspot.com/125540043/diff/80001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/80001/content/constants.article... content/constants.article:194: Integer constants default to `int`, floating-point constants `float64`, rune constants to `rune` (an alias for `int32`), imaginary constants to `complex128`, and so on. I don't think there's an "and so on". That's all the options. https://codereview.appspot.com/125540043/diff/80001/content/constants.article... content/constants.article:442: const milliSecond = time.Second/1e3 s/milliSecond/millisecond/. caps are for the beginning of a new word. millisecond is one word.

[r, 2014-08-22 18:51:13]

https://codereview.appspot.com/125540043/diff/80001/content/constants.article File content/constants.article (right): https://codereview.appspot.com/125540043/diff/80001/content/constants.article... content/constants.article:194: Integer constants default to `int`, floating-point constants `float64`, rune constants to `rune` (an alias for `int32`), imaginary constants to `complex128`, and so on. On 2014/08/22 00:39:36, rsc wrote: > I don't think there's an "and so on". That's all the options. Done. https://codereview.appspot.com/125540043/diff/80001/content/constants.article... content/constants.article:442: const milliSecond = time.Second/1e3 On 2014/08/22 00:39:36, rsc wrote: > s/milliSecond/millisecond/. > caps are for the beginning of a new word. millisecond is one word. Done.

[adg, 2014-08-25 01:44:08]

*** Submitted as https://code.google.com/p/go/source/detail?r=0362e1feb038&repo=blog *** go.blog: constants LGTM=gri, adg, rsc R=gri, adg, rsc CC=golang-codereviews https://codereview.appspot.com/125540043 Committer: Andrew Gerrand <adg@golang.org>