Add design for sql storage, implementation of db that stores the data, not retrieving yet.

Add design for sql storage, implementation of db that stores the data, not retrieving yet.

BUG=skia:

Committed: http://code.google.com/p/skia/source/detail?r=14175

[mtklein, 2014-04-11 18:12:40]

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/DESI...
File experimental/webtry/DESIGN.md (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/DESI...
experimental/webtry/DESIGN.md:142: code      TEXT      DEFAULT ''               
 NOT NULL,
Image at the time of creation too?  If these are for bugs, it'd be good to have
that as a reference point.

Or, dare I suggest it, another table to support multiple timestamped (+ other
meta) images?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/TODO.md
File experimental/webtry/TODO.md (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/TODO...
experimental/webtry/TODO.md:15: - Add in all the Sk header files that could be
needed.
Done?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/TODO...
experimental/webtry/TODO.md:16: - Add -port flag to webtry.go.
Done?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
File experimental/webtry/webtry.go (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:40: useChroot = flag.Bool("use_chroot", false,
"Run the compiled code in the schroot jail.")
Think we should change this to something like runLocally to control whether or
not we useChroot and whether or not failing to connect to the DB is fatal?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:75: req, err := http.NewRequest("GET",
"http://metadata/computeMetadata/v1/instance/attributes/password", nil)
https?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:210: code := string(b)
gofmt?  Seems some of this is tabs and some is spaces.  Either's good by me.

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:216: writeToDatabase(hash, code)
This has got me thinking about what should be going into that hash.  Just code,
code and resulting image?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:268:
log.Fatal(http.ListenAndServe(fmt.Sprintf(":%d", *port), nil))
An alternative I see a lot in Go code is to make port a string flag, defaulting
to ":8000".  This lets you also serve off a particular IP only if you want.

[jcgregorio, 2014-04-11 20:24:01]

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/DESI...
File experimental/webtry/DESIGN.md (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/DESI...
experimental/webtry/DESIGN.md:142: code      TEXT      DEFAULT ''               
 NOT NULL,
Yeah, that should be a separate table with the git hash + code hash as the
index. Even then, I'm not sure about storing the image in the database.

Regardless, nothing I want to tackle in this CL.

On 2014/04/11 18:12:40, mtklein wrote:
> Image at the time of creation too?  If these are for bugs, it'd be good to
have
> that as a reference point.
> 
> Or, dare I suggest it, another table to support multiple timestamped (+ other
> meta) images?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/TODO.md
File experimental/webtry/TODO.md (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/TODO...
experimental/webtry/TODO.md:15: - Add in all the Sk header files that could be
needed.
On 2014/04/11 18:12:40, mtklein wrote:
> Done?

Done.

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/TODO...
experimental/webtry/TODO.md:16: - Add -port flag to webtry.go.
On 2014/04/11 18:12:40, mtklein wrote:
> Done?

Done.

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
File experimental/webtry/webtry.go (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:40: useChroot = flag.Bool("use_chroot", false,
"Run the compiled code in the schroot jail.")
On 2014/04/11 18:12:40, mtklein wrote:
> Think we should change this to something like runLocally to control whether or
> not we useChroot and whether or not failing to connect to the DB is fatal?

I actually prefer finer grained control. For example, to debug a problem I might
need to run a second instance on the server on port 8001, but not run in the
schroot so that I can debug the code.

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:75: req, err := http.NewRequest("GET",
"http://metadata/computeMetadata/v1/instance/attributes/password", nil)
The metadata server is http only.

On 2014/04/11 18:12:40, mtklein wrote:
> https?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:210: code := string(b)
I've got my editor to run go fmt on Go files every time I save. This seems like
a weirdism in Rietveld, as the entire file is tabs.

On 2014/04/11 18:12:40, mtklein wrote:
> gofmt?  Seems some of this is tabs and some is spaces.  Either's good by me.

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:216: writeToDatabase(hash, code)
See previous comments on saving the image in the database.

On 2014/04/11 18:12:40, mtklein wrote:
> This has got me thinking about what should be going into that hash.  Just
code,
> code and resulting image?

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:268:
log.Fatal(http.ListenAndServe(fmt.Sprintf(":%d", *port), nil))
On 2014/04/11 18:12:40, mtklein wrote:
> An alternative I see a lot in Go code is to make port a string flag,
defaulting
> to ":8000".  This lets you also serve off a particular IP only if you want.

Done.

[mtklein, 2014-04-11 21:35:45]

lgtm

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
File experimental/webtry/webtry.go (right):

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:75: req, err := http.NewRequest("GET",
"http://metadata/computeMetadata/v1/instance/attributes/password", nil)
Doesn't that expose the DB password then?

On 2014/04/11 20:24:01, jcgregorio wrote:
> The metadata server is http only.
> 
> On 2014/04/11 18:12:40, mtklein wrote:
> > https?
>

https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
experimental/webtry/webtry.go:210: code := string(b)
On 2014/04/11 20:24:01, jcgregorio wrote:
> I've got my editor to run go fmt on Go files every time I save. This seems
like
> a weirdism in Rietveld, as the entire file is tabs.
> 
> On 2014/04/11 18:12:40, mtklein wrote:
> > gofmt?  Seems some of this is tabs and some is spaces.  Either's good by me.
> 
Ah.  Good.  I must have just been confused.

[jcgregorio, 2014-04-12 02:24:51]

On 2014/04/11 21:35:45, mtklein wrote:
> lgtm
> 
>
https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
> File experimental/webtry/webtry.go (right):
> 
>
https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
> experimental/webtry/webtry.go:75: req, err := http.NewRequest("GET",
> "http://metadata/computeMetadata/v1/instance/attributes/password", nil)
> Doesn't that expose the DB password then?

No, the metadata server is only accessible to VMs, not outside traffic. Also,
the metadata server only returns the data for an instance based on which
instance is
requesting it.

> 
> On 2014/04/11 20:24:01, jcgregorio wrote:
> > The metadata server is http only.
> > 
> > On 2014/04/11 18:12:40, mtklein wrote:
> > > https?
> >
> 
>
https://codereview.chromium.org/235373002/diff/40001/experimental/webtry/webt...
> experimental/webtry/webtry.go:210: code := string(b)
> On 2014/04/11 20:24:01, jcgregorio wrote:
> > I've got my editor to run go fmt on Go files every time I save. This seems
> like
> > a weirdism in Rietveld, as the entire file is tabs.
> > 
> > On 2014/04/11 18:12:40, mtklein wrote:
> > > gofmt?  Seems some of this is tabs and some is spaces.  Either's good by
me.
> > 
> Ah.  Good.  I must have just been confused.

[jcgregorio, 2014-04-12 02:25:00]

The CQ bit was checked by jcgregorio@google.com

[commit-bot: I haz the power, 2014-04-12 02:25:14]

CQ is trying da patch. Follow status at
https://skia-tree-status.appspot.com/cq/jcgregorio@google.com/235373002/80001

[commit-bot: I haz the power, 2014-04-14 14:54:18]

Change committed as 14175