[Jack Gibson]

I think -merge m:m- should be eliminated, plain and simple. (...) it's just a dangerous trap for unsuspecting users.

I'd be equally happy with that option

[Belinda Foster]

One of the things i would like to see in a next version is more choice regarding symbols for line graphs (or the ability to assign our own). This is valuable when the same graph contains more than 4 or 5 lines and currently the alternatives all look very similar. Excel provides quite a few choices. I would also like to have the ability to easily assign patterns to bar graphs.

[Joseph Coveney]

One of the things i would like to see in a next version is more choice regarding symbols for line graphs (or the ability to assign our own). This is valuable when the same graph contains more than 4 or 5 lines and currently the alternatives all look very similar. Excel provides quite a few choices. I would also like to have the ability to easily assign patterns to bar graphs.

I'm not sure whether the following code snippet will render correctly your browser, but you can assign any symbol that can be expressed in Unicode and that is supported by the font that you're using.

sysuse auto
generate byte sequence = _n
label variable sequence Sequence
generate str marker1 = "▼"
generate str marker2 = "▲"
graph twoway ///
    connected mpg sequence, msymbol(none) mlabposition(0) mlabel(marker1) || ///
    connected turn sequence, msymbol(none) mlabposition(0) mlabel(marker2) ///
    ylabel( , angle(horizontal) nogrid) legend(off)

In lieu of graphs that contain numerous plots, consider multiple simpler graphs.

Patterns in bar graphs has come up on the list from time to time over the years. You might want to search the archives to see what's been said on the topic and for recommended alternatives.

[Belinda Foster]

Unfortunately the only enlightening post i manage to find on patterns is that of Sergiy Radyakin but the implementation within programs appears difficult. The fact that there is no detailed documentation for the Stata graphics engine is a major obstacle as well.

[Joseph Coveney]

There might be one or a few other posts both as to why StataCorp might have so far declined to support fill patterns in bar graphs and as to some alternatives to consider.

[Nick Cox]

Joseph has given a link to a post of mine with my personal views on area patterns (just say No).

I've suggested to StataCorp the following extra marker symbols

pipe | (not a high priority as faking them with marker labels is easy)

triangles with apex down, facing left and facing right (to complement ms(T))

arrows facing in eight directions (N, NE, E, SE, S, SW, W, NW)

vees V (logical or in some notations)

wedges (vees upside down; logical and in some notation)

[Belinda Foster]

Nick Cox are there any sources of information regarding the Stata graphics engine? I would like to know more about it but it appears that there aren't any manuals, papers etc available. The only thing i have found is a presentation that Sergiy Radyakin had given a few years ago.

[Nick Cox]

What precisely do you mean by Stata graphics engine?

No user to my knowledge has gone deeper than Sergiy; equally I don't think Sergiy has documented all that he found out, or even that he added.

In essence StataCorp appears to have decided that the cost of writing as it were another manual was too great to bear. By my rough at-a-distance estimate it would have tied up about two developer-years and lots of other projects would have been delayed or cancelled.

My experience as a programmer of graphics extras is that whenever I think I need to know something about the graphics class system or something equally deep I eventually discover that there is a twoway trick to get round it.

[Garret Christensen]

I'd love to see some incorporation of version control into the do-file editor. I use Git, but flexibility to use Git or SVN would be nice. R Studio has it built in, and it's quite convenient.

[Sergiy Radyakin]

Garret, you can use version control right now. And that's what many users are doing.

The problem that I see is that Stata's project file *.stpr is binary, so resolving versioning conflicts is not straightforward, compared to it being a plain text, ini, or xml file.
It would be best if we could have it in some sort of text-based format. For example, compare with MS Visual Studio project file format.

Thank you, Sergiy Radyakin

[Jimmy Yang]

I wish

tabulate varname, sort

can sort not just frequency, but also by variable's value.
it's not friendly to use

svmat

to save result for further inspect.
since tabulate is such an important tool, it'd better to provide more option.
And related issue, please expand the limit of

encode str

I hate to use

egen varname=group(str)

why put a limit on encode, I don't understand?

[Garret Christensen]

Sergiy Radyakin I could be mistaken, but I think that Stata's project management (which I didn't know about) and version control are quite different. The project manager will let you keep track of which files go with which filepaths, but version control is a lot of different capabilities--most importantly keeping track of changes to your files.

[Sergiy Radyakin]

Garret, they are different and one does not preclude the other.

I wrote:"you can use version control right now. And that's what many users are doing.". I am using Mercurial now, used Bazaar and SVN in the past.
You wrote you are using Git. They are all perfectly viable options with Stata.

Why integrate the version control into Stata itself? Tortoise provides an excellent GUI client already.

On the other hand when you are using these tools now, and if you were using Stata's already existing projects (*.stpr), then you encounter the problems with conflict resolution based on the binary project where you can pick left or pick right, but not get a bit of left and right.

Sergiy

[Marco Savegnago]

Dear all,

after

generate newvar = expression [if] [in]

it would be nice to have an automatic label to the newvar containing the expression (and the possible conditions).

So in the example ,

webuse genxmpl3, clear 
generate int age2 = age^2 if age > 30

the "new" generate could actually add (in the background)

label variable age2 " = age^2 if age > 30 "

This feature could be switched off with a nolabel option.

Best,

Marco

[Nick Cox]

Marco:

search defv