Figures in LaTeX documents (what to use?)

I used affinity not long ago to edit a product label for a local company. It wasn't too hard to get it to do what I wanted but I didn't have to create a new complex diagram from scratch.

Depending on the type of diagram you want there's also Visio which is not officially free but I don't think Microsoft care any more as long as you're not a big business user of many copies.

A lot depends on how complex the diagrams will be and how much you will need to tweak your work while creating them.

E.g., ideally, you want to build a diagram out of "drawing objects" and not just "FLATTENED marks on paper". The former allows you to process those objects even after they have been placed (e.g., to move them, resize them, change color/typeface, etc.) The latter is like using a whiteboard -- tomake a change, you have to carefully erase the old (while not disturbing anything that was drawn over or under it!) and redrawing anew.

Where possible, I like to rely on photographs that I can annotate with callouts, highlights, etc. (easier to get things right!). In the past, when getting photographs wasn't convenient, I would build

If the thing being drawn is a physical device, I like to use CAD as it lets me define relationships between the various parts of the drawing in "real units" instead of eyeballing things (e.g., "this part is

For certain *types* of drawings, you can find tools that are designed with those concerns in mind. E.g., when I am documenting state machines, I need to move things such that connections are maintained (rubber-banding).

For graphs and other content that is mathematically defined, I use tools that will render those relationships *for* me.

Having said these things, I use:

- Adobe Illustrator

- CorelDesigner

- Adobe Photoshop

- Corel PHOTO-PAINT

- gnuplot

- Mathematica

- FontLab (to create "fonts")

- Metashape (SfM)

Note all of these predate cloud hosting. I have no idea what's available in current offerings.

Consider, also, addressing annotations "above" the notion of the illustration (i.e., in the publishing tool). This lets you map the typeface to that used in the rest of the document, better (I've embedded Mathematic charts in publications and the difference in "fonts" stands out noticeably.)

It also lets you scale and translate the drawing without also affecting the annotations that may be *in* the drawing.

Use Word! You can drag-and-drop pics into a document. And export a PDF.

I sometimes draw on paper and scan it. Then I don't have to fight the computer for exactly how I want it drawn.

Yes, we do that, and put pics of whiteboards in docs. But Certain Parties think that stuff that customers should see, should look more grown-up.

I use Visio for some figures. One can also screen cap LT Spice screens, or sheets of a schematic; all our schematics have a sheet 1 block diagram.

Photos of the hardware are good in manuals too.

My guys use some AI thing to tweak photos. Change text, change connector types, remove dust.

Do you want diagrams as 'code' or draw them in a GUI?

For the 'code' approach, TikZ is popular these days. You may find an addon package that draws just the kind of diagram you want, in which case it should be easy to get 'something', and harder to get exactly what you want (just like Latex really). If you have a lot of diagrams of a similar type this might be the way to go, but if you need to hand edit them it can get a bit annoying. eg CircuiTikZ is a package for drawing schematics.

There are tools which let you edit the source code on one side and continually recompile it so you can see the output on the other side - eg this is a web app:

For the 'drawing' approach, I've used Inkscape and have it emit PDFs which you can \includegraphics. It's free and open source. It uses SVG natively and I think some versions of Latex can import SVG, but PDF is less finnicky. It works similarly to Adobe Illustrator and other vector drawing packages.

There's some tools which will emit both a diagram file and a Latex file with the text. These are nice in that the text is in a matching font and size to your body text. I think matplotlib / PGFPlots does that, but that's more for graphs than arbitrary figures. I'm sure I came across a flow for doing that for general drawings, but I can't remember where it was.

Theo

Absolutely, but I think all drawing software now works like that.

Yes, that's what I'm looking for. lpe does that, and I see the inkscape can also do this.

I would like to draw them.

Yes, I see that TikZ has (as I understand it) taken over from PicTeX and PStricks. I have been browsing CTAN and don't recognise 99% of the packages - but it's good that LaTeX is still so popular.

Thanks for the Inkscape pointer, I see that it has a PDF+latex save option, which does exactly what you describe, and is the sort of thing I was after.

There's a short demo of it in this youtube video

Time is money, so I try and use the best tool(s) for the job in hand.

I'm going to give it a try, maybe not for this work, but it does seem to be a no brainer give it a go.

Whenever I have tried to do something slightly unusual and tried a variety of packages (that might be easier to use) I always ended up using either inkscape or GIMP depending on whether it was a bitmapped or vector requirement. John

This actually can make your job *harder* as it isn't always easy to "pick" which object you want to interact with -- when objects are located close together. E.g., a text object inside a circle object may be tedious to access as the circle may grab the focus, instead. This because the pick/select tool isn't selective. You may have to "lock" one object in order to keep it from being selected as you try to pick another nearby/within.

Photographs also have the advantage of giving you an opportunity to bring large splashes of color into a document. This *really* dresses it up -- moreso that colored line segments, etc.

The Adobe tools have the hidden advantage that you can edit the text representation of the drawings as it is usually something like embedded postscript. So, if you wanted the ends of several linesegments to begin at the same "X" position on the page, you can manually tweek those values in the drawing's representation because you often can't position the mouse cursor in exactly the same location when drawing another object.

If so, then you want to avoid making those annotations in the drawing itself.

You might also want to consider if the tool you select supports things like footnotes and cross references. I find that I use both mechanisms heavily as it lets me take annecdotal text out of the body of the document yet keep it available to the reader.

I also use color as an aid to myself in verifying that I have tagged specific text/paragraphs with the correct style tags. E.g., foreign words like /etc/, /ibid/, /et al./ should always be italicized. But, you can manually apply italics to selected text. *OR*, you can create a text style "foreign" and apply that, instead. In the latter case, you can change how "foreign"-tagged text is presented simply by changing the style definition. In the former case, you have to go through and manually change/remove the italics that was applied.

Knowing what tools can do and their representations is a big asset.

When I had to create the default templates for my gesture recognizer, I relied on Illustrator to actually "compose" them as I could ensure certain desirable geometric relationships that I couldn't guarantee freehand (e.g., that the top of a '2' was exactly as wide as the bottom).

Then, extracted the representation of the drawing from the .AI file to create an all-bezier representation for my software. Then, rendered these in gnuplot to be imported to the user manual.

Used Visio for some years, an older version from Ebay. However, Libre Office, (from Cern, os agnostic: linux, windows and more, has a good drawing package which will import Visio and other formats for re editing.

Small learning curve, but worth the effort.

Chris

I also use an old Visio. It's OK and if you export an EMF file, it does very nice line work in a Word doc.

I have a sheet of electronic symbols if anyone is interested.

The whole idea here, long term is to get rid of everything from Microsoft. There has been an ongoing effort for years at Cern,

I endorse geting rid of Microsoft garbage, but I think that Linux is ancient tradition too. We need a new idea.

Linux is 34 years old and is basically Unix which is is 57 years old.

Pixellated graphics, especially screen captures, never looks as good as vector stuff.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics

Pure chronological chauvinism. I’m older than Unix, and so are many other fairly up-to-the-minute folks.

My beautiful line drawings are usually made with Freelance 4.0 for DOS. They come out in native Postscript, which works great in LaTeX, especially pdflatex.

Cheers

Phil Hobbs

Perhaps, but constantly updated and improved, though ever since systemd, and the red hat / ibm takeover, have been running FreeBSD for lab server and software dev machines. Xubuntu Linux and Suse are here when needed, but not that often. Many years on Sun / Solaris as well, excellent and rock solid back in the day, but outclassed now.

Linux may be old, but the original ideas that inspired unix still apply in the various open source variants, if not in Linux.

Chris

CPUs used to be expensive so had to be shared and later threaded among both the bits of the OS and among user apps.

Times have changed. Basic OS architectures haven't.

John Larkin Highland Tech Glen Canyon Design Center Lunatic Fringe Electronics