Wikipedia talk:WikiProject Chemistry/Structure drawing workgroup/Archive May 2007

The size of the images

What's the point of creating and uploading THIS Image:1,3-bisphosphoglycerate.png kind of image? The drawing style i use is - bond-length: 0.24in; bond-width: 0.7pt; font (main text): 9pt, Arial, bold; font (super- and sub-scripted text): 7pt Arial normal. and i'm trying to use it with ChemSketch as well. -- Boris 04:20, 10 April 2006 (UTC)[reply]

OK, let's focus on single structure images for now. Boris, can I take it that you're saying we don't need to upload huge files?! I agree, but I think equally well drawings like Image:Prostaglandin I2.png and Image:Cysteamine.png are too small. If someone wanted to use these drawings on the front page of their term paper, or even just on a different page at double its size, it would be way too small. I have seen this problem in the past. For example in the formic acid page, if we wanted to get a reasonably prominent image of the structure we couldn't in version. I think we need a happy medium. I have been told that wiki software scales images down automatically - so if you have a 100 k image showing on a page at one quarter of its original area it only takes 25k of downloading - can anyone absolutely confirm this? If this does not happen we need to limit the size of things more strictly.

In a recent image I created in ChemDraw and posted, Image:Quassin.png, I used 1.058 cm (0.4167 in or 30 pt) bond lengths, 1pt bond widths, 14pt Arial font for atom labels. I'm not wedded to those settings, they were derived from the settings given in the style guide by scaling down in size. I'd be happy to see the font a little bigger or the bond lengths a little smaller for when the image is shown smaller, as on the quassin page. I then saved it as a TIFF file (this image came out as 1037 x 764. I then opened it in IrfanView, added a 75px margin as recommended in our style guide, and saved it at 593 x 457px (this is probably half the size of the original).

Regarding IsisDraw, I can't get the "Export to TIFF" (or to JPG) in IsisDraw 2.5 to give me a file readable by IrfanView. Last year, before I got ChemDraw, I made my pics in IsisDraw then scanned them in, for an example see Image:Menthol structures.png. For this I simply used the default settings for IsisDraw, with one exception - go to Options-Settings-Chemical Drawing and on "Show hydrogen labels" select "on hetero" in order to make sure O-H and NH2 groups don't appear as just O and N (Boris, you need to fix this in the cysteamine pic, IMHO). I scanned the image in at (I think) 300 x 300 dpi, then scaled it down in size to get the image shown. I like the proportions of that image better, but obviously a scanned image is going to have poorer resolution - though this one doesn't look too bad.

Aren't there ways to print to a virtual printer that outputs a digital image file instead of having to physically print and scan the image? -- midorigin 15:50, 9 April 2007 (UTC)[reply]

What do others use for their settings for simple compound structures in ChemDraw etc? Any comments on the above? Walkerma 16:16, 10 April 2006 (UTC)[reply]

Some time ago I wrote a small tutorial where I listed the settings I use to create simple structures in Chemdraw. See it here! -- Run e Welsh | ταλκ 17:17, 10 April 2006 (UTC)[reply]

I use RuneWelsh's settings (or something very similar). I think they produce nice images - most of them need to be scaled down in the articles, but having large images available is good, in my opinion, because it allows for a wider range of uses than smaller ones. --Ed (Edgar181) 17:29, 10 April 2006 (UTC)[reply]

I use a modified version of Cacycle's settings:

Line width: 0.107 cm
Bold width: 0.4 cm
Hash spacing: 0.3 cm
Bond length: 2.566 cm
Font: Arial 45 pt
Copy to clipboard and paste into Photoshop (or save as TIFF black/white bitmap if clipboard method fails)
Image processing (Photoshop Elements 3.0):
Resize image to 1000px width
Add canvas of 50 pixels to all four sides (Photoshop: Image:Canvas Size, relative, 100 pixels width, 100 pixels height)
Save as a PNG black/white bitmap

I sometimes modify some of these settings to suit an individual image, usually hash spacing to make sure there are enough hashes in a hashed bond (eg. more than two) so you can tell it is actually a hashed bond!

I am, however, happy to use whichever settings we decide are preferred.

As for image size, what's wrong with uploading large images? You can resize them to whatever size you like. I haven't noticed Wiki taking longer to resize larger images. Maybe it does though. I don't think colouring atoms by element in structures is a good idea - it's redundant and is hard to read. Yellow against a white background for sulfur is hard to see. Fine for 3D though, as the shadow effects make yellow perfectly visible against white. With ISIS/Draw, can't you just copy to clipboard and post in Photoshop or IrfanView?

Happy drawing! --Ben 17:38, 10 April 2006 (UTC).[reply]

As far as I know the images produced using the settings I linked above tend to render rather nicely when they are resized automatically by the wiki software. Large images also have the advantage that they can be manipulated more easily to look nice on print. -- Run e Welsh | ταλκ 19:02, 10 April 2006 (UTC)[reply]

Ben, you're right. Even though there's not an export option in ISIS, you can copy/paste to a graphic editor (including Window's Paint, which is not the best choice, of course). I also verified that the size of the image doesn't depend on the zoom level, so it's a reliable option. But be aware it adds a 6px canvas on each side (no matter what size the image is). This copy/paste mode also works for ChemDraw, but the canvas is 3px this time.

I've only uploaded a couple of images using Rune.welsh's version of ca Cacycle's settings, very similar to those used by Ben. I guess they are OK to most purposes. I agree with Ben about resizing hashed bonds when needed, but it's a minor issue. I do think it is important to keep the images big. I think it's not such a great difference in file size, but the images render nicely when resized by Wikipedia, with a beautiful antialias, and they can be used elsewhere without worrying about pixelation. Finally, I find it really important to show hydrogens in heteroatoms, it should be a standard. --Nevermore78 20:39, 10 April 2006 (UTC)[reply]

Walkerma, I’ll go with your idea for the size - here is an example (600X335), which could be made to be exactly twice (640X357) the original size (320X179). Why I use (320X179) you might ask? Well few years back when I started my own website I thought that this size could fit most of the structure and as I had already had quite a number of them I didn't want to resize the background rectangular that set the dimensions, b/c that resizing would had been a very time consuming work, so I stuck with this funny size. Anyway, I used to draw these kind of diagrams in black and white. Few months ago I started to color them because I think that the coloring engages the color processing centers in the brain and thus allows more information about the structure to be presented in the image and faster to be processed by our brains - in molecular biology when showing the surface charge distribution of a molecule, mostly proteins, the blue symbolizes the (+) charge and is almost exclusively associated with the N-atoms, while the red symbolizes (-) and is associated with O-atoms, the same goes for S, P, etc. all these atoms have their properties so the compound is expected to "behave" in a certain way where these atoms are present, and its a lot easier to spot their position in the structure when they are colored that when they all are in plain black. I have omitted the H-atoms because as I said before they don't provide any useful information except when they are at optical center such as those in the steroids, or the morphinanes, if the diagram has "-O" then that's a hidroxyl group, "-O-O" - peroxy, "=O" carbonyl, the same for N - "-N" primary amine, "=N" imine, "-N-" secondary amine, etc - we know the number of the hydrogens because we know the properties of the atoms and we have already agreed for no disociation states in the infobox images, if we can draw without the carbon hydrogens and we are used to it why not without the rest. I don't show the atoms at the terminal -CH₃ groups either, the way you have in Image:Quassin.png. The hydrogens are very important in the reaction mechanism diagrams though or in those "origin" images. So my drawings show all atoms but the hydrogens, and those that are not C are in colored text - simple with style.

I don't have 40 something inches screen so i will not use the "2.5 cm bond, 45pt font" settings just to have those "printer-friendly" big azz images. No way. But if i export the 320X179 diagram like this - "Export-->Other...-->JPEG" - it is converted into a 2397X1341 ".wmf" that can be downsized by a good Image editing program (InfanView does not give good results, i use DellPhotoEditor instead, but it's good for file convertion) as much as i want (see the begining of the reply) including all the way to the original dimensions or even smaller than that. -- Boris 02:52, 11 April 2006 (UTC)[reply]

Thanks for your explanations, Boris. We seem to have a majority of folks favouring large images, but this is not unanimous. I have posted a question at the village pump to help resolve the issue about automatic resizing of images once and for all. Personally, I think Rune Welsh's and Ben's settings look really nice on the Wikipedia page, and I would like us to standardise on something that looks like that. At the same time, I can understand Boris's complaints about gigantic images - I think a typical steroid would be too big to see on my screen while drawing - surely our pictures don't need to be that big? I wonder if we can come to some sort of compromise, mostly by scaling Rune Welsh/Ben's numbers by 2/3.

Chain angle 120
Bond spacing 15%
Fixed length 1.693 cm (48 pt)
Bold width 0.2 cm (5.67 pt)
Line width 0.071 cm (2 pt)
Margin width 0.106 cm (3 pt)
Hash spacing 0.071 cm (4 pt)
Arial 30pt as the Atom labels font
Times New Roman 30pt for captions, in case we need to add a small caption.

I altered the Quassin image to these settings in ChemDraw. I saved it as a TIFF file, then opened it in IrfanView. I resized to 600px width, put a 50px margin on all four sides, then saved it as a png file. The TIF file was 47k, and the PNG was 38k (I think IrfanView is not that efficient in making PNG files). Then I uploaded it, and you can see it on the right, click to see full size (it's still plenty big!). Yes, I know the methyl groups are messy, that's another issue, but I think the proportions look right, much the same as in Ben/Rune Welsh's. Would these settings work? Can we adapt them for the other software? Walkerma 19:18, 11 April 2006 (UTC)[reply]

How these settings look when used by Ben

What a negotiator! I'm going to start using these settings from now on, I'll post any feedback here. When I draw steroids or other large molecules in ChemDraw, I simply set the zoom to 50%. Then once I've copied the image to Photoshop, I resize it to 1000px or similar. But I can understand expansive structures being a problem during the production process, so to speak. I think we should definitely show hydrogens on heteroatoms - more than anything else, it's standard practice almost everywhere - we're not in the business of confusing people here! I actually quite like Boris' use of color atoms having looked at a few, but the use of pale colours is going to remain problematic, especially light grey for hydrogen. Cheers for your input guys, I'm enjoying this. --Ben 21:10, 11 April 2006 (UTC).[reply]

I just made my own version of quassin using the new settings. Everything is great, ~~except I think hash spacing should be about 0.18 cm, otherwise it doesn't look right with my set-up~~. On second thoughts, the hash spacing is probably OK. I'll mull over it. Great job Martin. --Ben 21:24, 11 April 2006 (UTC).[reply]

I saw the post on the Village Pump, so I thought I'd make a few comments here. Firstly, try not to make any guidelines based on technical concerns, such as file size. The most important issue here is the clarity of the images. Anything beyond that is just implementation, which doesn't really matter. Secondly, don't worry about the file sizes on the pages. The Mediawiki software uses ImageMagick to automatically scale the image on the server, so that when a page needs a thumbnail, it only gets a thumbnail (not the big image).

Having said that, PNGs such as these should be black and white or grayscale, non-interlaced, and have no transparency. This will produce the smallest file size. If anyone needs help with this, I am willing to compact the files (with no impact on quality)--just leave me a note on my talk page.

Ideally, these images should be produced as SVG images. SVGs are vector images, which basically means that instead of defining the pixels of the image, you define the shapes, so that even if you resize the image, it's still sharp (because the computer recalculates all the shapes when you view it). SVGs for these kind of images would be very, very small, and would look good no matter how big they were. Actual dimensions would not impact file size. However, I don't know what chemistry software packages support SVG (a quick Google search turned up JChemPaint), and I'm not sure how easy it is to draw them in Inkscape (a free SVG program I've actually used). ~MDD 46 96 21:43, 11 April 2006 (UTC)[reply]

SVG would be ideal. I've tried exporting from ChemDraw to SVG using a free program called SVG Factory but it doesn't render the element symbols (a bug, perhaps?). I can't afford Adobe Illustrator and Inkscape won't let me paste ChemDraw structures from the clipboard on my Mac. So that's the status with me and SVGs at the moment, but I'd love to use them if I can get the software working. --User:Benjah-bmm27 21:59, 11 April 2006 (UTC).[reply]

Agreed. SVG is perfect for the application. Unfortunately, the last time I looked (about a year ago) I found no good way to make SVG chemical structure diagrams.

ChemDraw does have an option to save in a native XML format that could be converted to SVG, if we had someone familiar enough with XML in general and SVG in particular who was willing to do the job. Shimmin 14:06, 12 April 2006 (UTC)[reply]

I also came here from the discussion at the Village Pump. I did a quick search on Google, and found a program called XDrawChem which is able to import the ChemDraw format (both binary and XML) and export to SVG. It might be worth a try. --cesarb 14:52, 12 April 2006 (UTC)[reply]

Another view on settings

I agree the above comments that large images are a Good Thing if they are PNG files, although this should not matter if we can get a SVG conversion running. The largest font size supported by ISIS/Draw is 48pt, so I scaled the RSC image recommendations to 48pt labels to get

Bond length: 83.7 pt
Stereo bond width: 11.0 pt
Down bond hashes: 12.3 pt
Line thickness: 3.4 pt

Even with these settings, the quassin is still completely within the editing screen, and it looks fine. ISIS/Draw will export as a WMF file, which should be convertible without problem to an SVG file, I will do some tests when I have installed SVG Factory. Physchim62 (talk) 15:19, 12 April 2006 (UTC)[reply]

I tried these settings in both IsisDraw and in ChemDraw. In Isis it's OK for something like quassin, but for some reason with ChemDraw it's completely unmanageable, it won't fit on my laptop screen. Could we choose something based on 48pt bond lengths instead, this will also ensure that Boris doesn't have a fit ("I don't have 40 something inches screen so I will not use the "2.5 cm bond, 45pt font" settings just to have those "printer-friendly" big azz images. No way.") Since our images are general much larger on the page than ACS/RSC images one might argue that our settings should reflect that - so we can use a slightly higher ratio of bond-length to font size? Could you live with that, PC?

If you want to alter the settings from those I proposed above to a "standard" style, perhaps we should base it on a scaled version of the ACS settings, simply because the ACS journals are high impact even in the UK. If you scale ACS settings by 3 you get 30pt atom labels and 43.2pt bond lengths, fairly similar to what I am proposing above, which in turn is based on the settings from Cacycle that most people are using. Walkerma 17:48, 12 April 2006 (UTC)[reply]

Well chemistry is an experimental science, so here are my preliminary results!

Another problem with large images is that ChemSketch users cannot change the font size or line thickness! Hence whatever is decided for ChemDraw users, a seperate set of parameters will have to be figured out for ChemSketch to get the closest similarity in images. As ChemSketch appears to use about 12 pt labels, I scaled the above parameters accordingly, which gives a 20.9 pt bond length (virtually the only modifiable parameter of interest in ChemSketch). This is obviously still too long, as you can see by the smaller size of the labels in the ChemSketch image as compared to the ISIS/Draw image: both images have been fixed at 200px width, as is standard for chemboxes.

I would note that there is no obvious difference in quality at the scale displayed on the WP image pages (the PNG files are at 300dpi resolution), dispite the fact that the ChemSketch image was exported as a TIFF bitmap whereas the ISIS/Draw image was exported as a Windows Metafile (vector graphics). Hence I'm tempted to agree that there is little to be gained in having large images. As for ACS vs RSC, I am completely neutral: I used the RSC recommendations as I knew where to find them!

Hint for ISIS/Draw users: if you cannot persuade ISIS/Draw to export a WMF, try exporting a TIFF: the TIFF will be unreadable, but you will also discover in the same directory a file $$$.wmf, which is in fact the WMF version of the structure. Unfortunately, SVG Factory will not convert it correctly into an SVG file, as mentioned above: the bug appears to be in SVG Factory. Physchim62 (talk) 12:57, 13 April 2006 (UTC)[reply]

Another issue to consider is that images that look best on screen may not be best in print. Personally, I prefer the image on the left above (because of the wider bond width, I think) on my screen. However, when I print them out, I think the one on the right is better. Similarly, ACS settings look good in print, but I prefer images generated with the current Wikiproject Chemistry guidelines on my screen. Since Wikipedia is mainly an online resource, we should focus on what is ideal for the screen and avoid defaulting to other common standards such as ACS settings, which have been optimized for their appearance on paper. --Ed (Edgar181) 13:16, 13 April 2006 (UTC)[reply]

Interesting, as it is the image on the left above which was created using settings optimized for print versions! I have tried to improve the line thickness on the ChemSketch image by shortening the bond length: the version with 15pt bonds is shown on the right here. I must say that bonds that short are difficult to handle on the screen while preparing the structure! Also, ChemSketch seems to adapt the label size for the bond length, so there are limits to what can be done to change the ratio of label to bond. It seems that we shall have to optimize first on ChemDraw and ISIS/Draw, then do our best for ChemSketch: we will never get complete compatibility. Physchim62 (talk) 14:18, 13 April 2006 (UTC)[reply]

Thanks for your input, PC. I'm glad you found the ChemSketch limitation, this is important. I also tested out my ChemDraw Quassin test image via WMF to PNG, and the PNG was about half the size of the one created via TIF. I also noticed a test $$$.wmf file after you mentioned that quirk of IsisDraw, that's good to know. Thanks, we should try to standardize on that if possible, I think. I tried Ben's suggestion of a simple cut & paste from Isis into IrfanView, and that worked, though the resolution wasn't great. I'm learning a lot here! Regarding Ed's comments, I reckon they are all pretty OK when printed off this page. If you print the full size image, I suspect the print appearance may be affected by the size - the thin lines of the right hand ChemSketch image are quite visible to the eye when full size, but begin to look faint when shrunk as on this page. The last ChemSketch sounds like it should look better, but IMHO it looks a lot worse, it even looks fuzzy at full size (on screen and printed). I'd much rather go with the original ChemSketch settings.

As for size, we need to balance the convenience of working with smaller images versus the benefits of larger images being useful for people to put in term papers, posters and Powerpoints. We also need to decide if we prefer to go with standard settings such as ACS, or whether we develop our own custom settings. But most users will be familiar with ACS, and both ChemDraw and IsisDraw come with ACS settings right out of the box. I was surprised to see that ACS settings do give quite a large sized image when converted to WMF, in fact both the PNG images end up as 15k. I also notice that there isn't a huge difference between these two (ACS vs ours) - not surprising, if Cacycle used ACS settings as a starting point.

If others can confirm this similarity of style, size and quality, I would propose that for simplicity we recommend use of standard ACS settings. That way any new chemist joining us can simply be told, "Use ACS settings, export to WMF then convert to PNG in IrfanView or Photoshop." Easy! For ChemSketch we use PCs original settings. Any comments, folks? Walkerma 05:59, 14 April 2006 (UTC)[reply]

ChemSketch also has built in ACS settings: the result is as shown! So I think the line is "Use ACS journal settings in any given molecule editor, export as WMF and use a image processing program to create a high resolution (300dpi) black-and-white PNG file. If you can create an SVG file, please do so and then tell us how you managed it!" This also means that molecular structures will be created at the same size for display purposes as for reaction schemes, which must be a Good Thing. Physchim62 (talk) 15:06, 14 April 2006 (UTC)[reply]

I have recently seen some problems with using the WMF file conversion from ChemDraw. I tried converting something at work that ended up completely mutilated (fonts squashed, far worse than K's example below). Also, I notice that the latest IrfanView has a "problem" (I just downloaded it, because my older version here at home couldn't add a canvas). Unlike the older versions, it reads ACS format files from ChemDraw or IsisDraw at actual size (i.e., tiny) rather than a much bigger size, whether WMF or TIF. I ended up using the old version if IrfanView to convert and the new version to add the canvas, you can see the result here. It is in use on the Chem portal news section. Is there something I'm doing wrong? Sorry I'm pretty much an amateur at this sort of thing, but perhaps I'm a good tester for the inexperienced newbie! Walkerma 03:26, 18 April 2006 (UTC)[reply]

Low-quality SVG images

I have noticed that certain users (including Bryan Derksen) replace chemical structures that exist as high-resolution PNG files with transparent background with low-quality and/or wrong SVG images (e.g. Image:Dimethyltryptamine.svg vs. Image:DMT.png) and tag the PNG version for deletion. This is clearly not scceptable and should be discussed first here. Please see also commons:Commons:Village_pump#Do_not_delete_.22superseded.22_images.21 and comment there.

Part of the problem might be that other users (including Stannered) tag chemical structures in guideline-compliant PNG structures with the template

This image was uploaded in the PNG, GIF, or JPEG format. However, it contains information that could be stored more efficiently and/or more accurately in the SVG format. If possible, please upload an SVG version of this image. After doing so, please replace all instances of the previous version throughout Wikipedia (noted under the "File links" header), tag the old version with {{Vector version available|NewImage.svg}}, and remove this tag. For more information, see Wikipedia:Preparing images for upload. For assistance with converting to SVG, please see the Graphics Lab.

without directly referencing here. Cacycle 15:11, 11 April 2007 (UTC)[reply]

Several things. Firstly, SVG graphics can be scaled up just as large as the PNGs with no loss of quality. Secondly, I haven't been marking the PNGs for deletion; the images need to be specifically marked for deletion in addition to being tagged as having a vector version available. I notice you've removed some of those vector version tags entirely, why are you doing that? And thirdly, I've been using BKchem which includes the chemical structure information within the XML file itself and thus gives them much superior editability over the PNG files. I don't see how this is lower quality. Bryan Derksen 17:04, 11 April 2007 (UTC)[reply]

The single removed a template stated that the PNG version should be deleted because a SVG version exists. However, the SVG version containes errors (it is missing an atom!) and its style is heavily flawed compared to the existing version.

I agree that in general SVG would be the more adequate file format. But the quality of a diagram or a structure does not depend on the file format, and, unfortunately, SVG is not yet supported by more professional structure editors (BTW, you can read all this at Wikipedia:WikiProject Chemistry/Structure drawing). Cacycle 02:26, 12 April 2007 (UTC)[reply]

As I've noted repeatedly, the superseded template is not actually a deletion template. I haven't been putting pngs up for deletion after uploading svg versions and there's no reason why they should be. Also, the "errors" you pointed out in this particular SVG were not actually errors but a variety of cosmetic details; this is not sufficient to declare the SVG image to be "flawed", it's just a slightly different presentation of the same data. I think you're overreacting. What are you proposing, that SVG versions shouldn't be posted at all unless they match your preferences? Bryan Derksen 05:33, 13 April 2007 (UTC)[reply]

Also I should note that this is an en.wikipedia guideline whereas I've been doing all this over on commons, a separate project that's used by all the various language Wikipedias. You probably shouldn't be trying to impose en guidelines on commons, instead I'd suggest taking this up over there. Bryan Derksen 17:14, 11 April 2007 (UTC)[reply]

We are talking about images that have been created primarily for use in Wikipedia articles and under the scope of this WikiProject. Additionally, we have considerable chemical competence bundled here. I would prefer that the Commons refer here for well thought-out guidelines on drawing chemical structures to prevent such 'accidents' in the future. Cacycle 02:26, 12 April 2007 (UTC)[reply]

For the record, I added the German style guidelines below (we need to be looking at these anyway) and I added an interwiki. As far as I can tell, no other Wikipedias have structure drawing guidelines. The German guidelines seem to say much the same as ours do, and I don't see any discussion (on the talk page) of problems with either png or svg. They don't seem to be pushing for replacement of png with svg. Walkerma 04:01, 12 April 2007 (UTC)[reply]

So continue using those PNG versions if those are the ones that are preferred here on Wikipedia. I have never proposed deleting any of them, I've just been providing SVG images of the same structures. I see no reason not to continue doing so. Bryan Derksen 05:33, 13 April 2007 (UTC)[reply]

OK, thanks! Hopefully the molecule editors like ChemDraw will start to support SVG! In the meantime we also have lots of things that don't have any drawing at all - take a look at this request list as well as these drug pages. Cheers, Walkerma 06:04, 13 April 2007 (UTC)[reply]

German style guidelines

My schoolboy German can't handle this, so I got Google to translate. It looks like they pretty much say the same as us, but they (a) advocate Public Domain for all simple drawings and (b) they only include terminal methyls when highlighting them (e.g. for stereochemistry). This is what Google gave me for the main graphics sections:

Recommended formats and upload

For structural formulas SVG or png are the recommended formats. Pictures in the png format should have a high resolution (width of at least 2000 pixels), in order to be able to be used in printable media. When such a high resolution is used, additional smoothing is unnecessary. Simple black and white is sufficient for 2D formulae. Uploading to Wikimedia Commons makes it possible for sister projects to use the structural formulas as well. From it follows also that connecting names or other German-language text should be placed in the picture legend, and not in the image file.

Licensing

Since structural formulas do not achieve the necessary creation height, in order to be protected, it is enough to publish them under a Public Domain license. Explanation concerning licensing [work on] Licenses are there to protect intellectual property. Most structural formula diagrams might, however, not constitute patentable works due to the small and/or missing creation height anyway (see §§ 1 and 2 exp. 2 UrhG), would be thus in common-free and could not at all be protected. Such structural formula diagrams should therefore not be placed under a license (neither the GNU FDL nor any CC license), but rather be marked directly as public domain (in Commons, with the tag {{PD-ineligible}}).

Uniformity of 2D formulae
Size

In order to reach within the Wikipedia a uniform appearance of the structural formulas, Heteroatoms should be use the font Arial with character size 13 pt, and bond lengths of 0,7 cm. In the somewhat better molecular editors these parameters can be adjusted manually. The accompanying picture can serve (in original size) as a comparison standard. An edge of for instance 5 px is recommended. Png files should possess a dissolution from 1000 to 2000 pixels. Smaller files are unsuitable for the pressure! With SVG the dissolution does not play a role.

Stand representation/detailed representation

Carbon atoms are usually not explicitly written, except if the atom is special — such as a chiral carbon, or as a "reaction partner", or for a chain with less than ten carbons. Therefore, terminal methyl groups are not represented as –CH₃, but as simple lines. Such a representation is known as a "skeletal formula", and it is the standard for us, except in the aforementioned cases. The Skeletal formula article provides further information on this type of representation.

Comments on German guidelines

Nice find, can you give the German reference? The Google translation doesn't seem bad at all, given the circumstances, but it would be nice to be able to link through. Two comments:

The guidelines make the comment about international usefulness, which is why we should not use "Me" for a methyl group (it can mean "metal" in German). The comments about text should also apply on English Wikipedia: apart from anything else, image text resizes badly (much less readable than atom labels).
Please do not upload anything but the simplest structure diagrams as {{PD-ineligible}}! The German guidelines give a correct statement of German copyright law, which is notorious for requiring a high level of creativity to qualify for copyright: in general (esp. in UK and US), structure diagrams should be assumed to be protected by copyright and hence a tag such as {{PD-user}} or {{GFDL-user}} should be used.

Physchim62 (talk) 17:36, 12 April 2007 (UTC)[reply]

Link is here, and also provides for 3D models. Myself and Rifleman have improved the original Google output, BTW. Fvasconcellos 17:49, 12 April 2007 (UTC)[reply]

I believe we had an inconclusive discussion about the copyright status of chemical structures some time ago. Somewhere. Anyway, I'm no expert, but what would amount to be the "simplest" structure? I've tagged *all* my original uploads of structures as pd-inelig. There are probably only a few irreducible representations of a molecule, and I don't think you can copyright that? --Rifleman 82 18:29, 12 April 2007 (UTC)[reply]

Harmonization of standards

I believe the en and de wikipedias are responsible for most of the structures drawn on commons. Our standards are essentially similar; can we use this as a basis for standards on commons, which are to be enforced for consistency? Ben was trying to organize the molecules at commons, but I think it's been quite a solo effort so far. Anybody an admin at commons? Any comments? --Rifleman 82 18:33, 12 April 2007 (UTC)[reply]

I had a chat with some of the Commons admins on IRC. They pointed out that we can't exactly stop people uploading structures, but otherwise they seem fine with it. Image deletion would have to be through their normal process, using the {{superseded}} tag. They suggest:

uploading a copy of the image guidelines to Commons as well, so that Commons admins can find them easily
creating tags such as {{chemstructure nonstandard}} for images which need replacing and {{chemstructure checked}} for images which have been checked and are OK.

Having guidelines on Commons should also help to avoid the problems with low quality SVG replacements... Physchim62 (talk) 22:06, 12 April 2007 (UTC)[reply]

Selfish question: would mine meet these standards? :D I don't use these exact editor settings, and tag all my uploads with {{PD-self}}. Fvasconcellos 22:32, 12 April 2007 (UTC)[reply]

They look fine to me—nice to know that someone can get ChemSketch and Inkscape to work togther to produce decent SVGs :) Physchim62 (talk) 22:40, 12 April 2007 (UTC)[reply]

:) Fvasconcellos 22:51, 12 April 2007 (UTC)[reply]

No need to worry, they are great :-) The problem seems not so much what exact setting is used but more the experience of a contributor in using as well as drawing chemical structures as well as certain background of chemical knowledge. We might want to prominently highlight our structure drawing service on the guideline page. Cacycle 23:16, 12 April 2007 (UTC)[reply]

Done :) Physchim62 (talk) 03:48, 13 April 2007 (UTC)[reply]

Wow! That's fantastic - thanks PC et al.! Walkerma 04:13, 13 April 2007 (UTC)[reply]

Our standards do seem very close to the German standards. Sounds like we should get their input on the ChemSketch project so both projects will feel confortable using the new ChemSketch wiki template when it is ready and tested !? -- Quantockgoblin 17:42, 13 April 2007 (UTC)[reply]

I already contacted them, I meant to mention it here but I was getting very tired...! Only a comment about BKChem so far - should we contact their main WikiProject to get more response? If so, could someone with better German than myself do this? Walkerma 17:55, 13 April 2007 (UTC)[reply]

(unident) Sorry my De is not up to scratch ... however, I bet there En will be pretty good! -- Quantockgoblin 18:05, 13 April 2007 (UTC) [reply]

Structure searching within Wikipedia

"Regarding the websearch add-in at http://www.acdlabs.com/servlets/UserAuth?pr=chemsk_websearch we COULD make Wikipedia structure searchable TOO. All we need to know is how to search it with a structure if its feasible."

Qgoblin, it looks like you have started things moving, towards making Wikipedia structure-searchable! Thanks a lot! This is quite a major issue, and a distinct topic by itself. that's why I started a separate header. The question is raised in the email - how can we make our chemicals pages (and maybe our reaction schems too) structure-searchable? It looks like InChI is the way to do that - is that right? - but how should that be done?

I think we need to make sure that whatever we tell ChemSketch now is something that allows things to develop. We don't want to find ourselves limited, then asking them to change the way they search our pages. Here are three stages I can see. We are already set up for Stage 1, but Stage 3 is much harder (but also more powerful. Do we want them to find just the article on the chemical compound drawn, or do we want all structures & reaction schemes to be findable?

Stage 1: We could try to add InChI into all ChemBoxes - that would be a start. Many already have it anyway. Is that all we can do for now? We 'd still need to make sure that someone searching on (R)-butan-2-ol would find 2-Butanol.
Stage 2: What would be ideal IMHO would be to draw an reaction scheme or structure in ChemSketch, and when we save it in Wikipmedia Commons it generates the InChI(s) and attaches it (them) to the file somehow. Can this be done? We want someone to be able to draw a structure in ChemSketch (or ChemDraw) and find it in a reaction scheme in the Commons.
Stage 3: Would we prefer for people to find the image file in context (i.e., in the Wikipedia article where the image is actually used), rather than just the isolated image file? I would say yes, but that would be even harder. I think we'd like to be able to type something like [[Image:MethylBenzoateNitration.png|300px|Nitration of methyl benzoate|reactant=1/C8H8O2/c1-10-8(9)7-5-3-2-4-6-7/h2-6H,1H3| product=1/C8H7NO4/c1-13-8(10)6-3-2-4-7(5-6)9(11)12/h2-5H,1H3]], but I guess that would need some changes to the Wikimedia software. Or is there a better way to do this? Walkerma 05:31, 5 April 2007 (UTC)[reply]

I believe that InChI is the best way to go at present for EXACT structure searches since it will be a string lookup. However, in order to integrate the exact structure search out of an application like ChemSketch there might need to be appropriate call to an Index of the strings. Is there any defined API for Wikipedia available for review or someone to talk to from the technical team to discuss the issue of doing this search? BTW, I am the product manager for ChemSketch and fully support doing this and will do everything in my power to make this work for the Wikipedia community --AntonyWilliams 14:58, 5 April 2007 (UTC)[reply]

I think it's great that ACDLabs is giving us a hand here. But I think when (or if) we are able to add in-line JMOL or CHIME-type files, all these silly issues about how best to present a chemical structure will be solved. I remember some rumblings of it then. Not sure by whom. Any updates? --Rifleman 82 15:29, 5 April 2007 (UTC)[reply]

I had the pleasure of a long chat in Chicago with Bob Hanson, the main writer of Jmol. They had some bugs which were solved about two weeks back. They are currently talking with people at Wikimedia (Tim Starling, I think) and Jmol should be part of Wikipedia very soon. I don't think Jmol is that good for structure searches, though, is it? Walkerma 15:43, 5 April 2007 (UTC)[reply]

Jmol is a web-based structure viewing tool as is CHIME. Structure searching is not dependent on these tools...viewing the results of these searches CAN use these tools. The structure searching is likely best facilitated with a synonym -based search (and we will try and set this up from ChemSpider or, as suggested, via InChI. I now have a beta version of ACD/ChemSKetch supporting PNG and will work with one of this group to facilitate testing by this group if there is interest. These conversations have started.--AntonyWilliams 17:14, 13 April 2007 (UTC)[reply]

That's very good news! In the meantime, I have been trying to get help on the technical aspects of adding metadata to structure image files, for example see this request. I'd like us to do this the best way possible. I've also been reflecting, I suspect that people might want to use this tool to post structures on their own websites, etc, not just Wikipedia. Will this tool be usable as a general "Post to the Internet" tool? Thanks, Walkerma 03:34, 15 April 2007 (UTC)[reply]

The API for MediaWiki is here. I too am trying to find the best technical person to talk to about this ('cos it certainly isn't me! :) Physchim62 (talk) 03:57, 15 April 2007 (UTC)[reply]

Hold on Martin, you are talking about making our images searchable (you have a reply at Commons, BTW), whereas I was thinking along the lines of making articles (more) searchable. Both are worthy aims, and the situation with articles is probably technically simpler, although we need to talk nicely to the developers to be sure, and to consider server-load from any solutions. An automagically-updated InChI index would seem technically feasible to me, at least in the medium term, but then I know nothing except that I know nothing! Physchim62 (talk) 04:30, 15 April 2007 (UTC)[reply]

Yes, I was trying to reach "Stage 2" on my above list. What I'd like to see ideally is

*You draw the structure in ChemSketch

You convert the structure to a "Wikipedia-friendly PNG" which has the InChI generated & attached as metadata.
You upload the file to Commons, and in the process the InChI is automatically loaded there too.
You put the image into an article, and the InChI is part of the readable metadata attached to that.

Thereafter, if you search for the InChI on the Web using Google (or from ChemSketch???), you would see the Image file come up as a "hit", and maybe the articles too. If you search within Wikipedia, for the InChI, you could find any articles that include that InChI. untimately I'd like us to reach Stage 3, where the Image parameters could include tags for reactant, product, etc., but that would have to be added manually, wouldn't it? (Assuming we could even get image tags like that created!). What did you specifically have in mind? As for Socrates, I also know nothing, I just ask the questions! Walkerma 05:30, 15 April 2007 (UTC)[reply]

Reactant tags would be great but even for SciFinder, their reaction assignments aren't all THAT reliable. I Tagging reactants before searching still gives me incorrect hits. --Rifleman 82 05:35, 15 April 2007 (UTC)[reply]

(reset indent) I think that the best way is for the InChI to be embedded in the image automatically by the molecule editor. We all know the hassle of hand-typing references. We'll have the same problem here. If it's too difficult, people will just NOT do it. Also, hand-typing it will give transcription errors and for such a complicated string, the probability of errors is very high! --Rifleman 82 05:23, 15 April 2007 (UTC)[reply]

My colleagues have investigated embedding data into the PNG file. For sure its feasible. We have taken a similar approach before when we embedded tags into PDF files generated by ACD/ChemSKetch about 4 years ago to make them structure searchable. This work will start shortly to see whether it can ber done in the near term to help this group move forward. Related question... I have a personal project (non- ACD/Labs related) at www.chemspider.com ChemSpider. We can embed in that system the ability to export SVG files following identification of a structure in the database. Then this group would simply search by trade name, synonym, registry number(s), retrieve a structure and export the file as SVG. I can put a statement on the website transferring copyright of exported SVGs to people doing this. Is this of interest and value do you think? --AntonyWilliams 15:27, 19 April 2007 (UTC)[reply]

That's excellent news! Be sure to look at this detailed discussion on the issue as well, this will help to see how things are handled from the Wikimedia end of things. It seems like the key issue is making sure that Google et al can find things like an InChI. I'm not sure about whether an RDF block, for example, can get picked up by Google, but I'm also very ignorant of this sort of thing, and you are a leading expert! Henry Rzepa (User:Rzepa) may also be able to help, he addresses some of these issues in his recent article. Many thanks, Walkerma 15:40, 19 April 2007 (UTC)[reply]

An update regarding the situation with embedding of InChI strings into a PNG file. The work has been done in the ACD/Labs development office to deliver this capability. Next we will make this a template inside ACD/ChemSketch freeware. I need to get one or more of this group to test this capability and provide feedback at your convenience. ALSO, ChemSpider has been linked up to Wikipedia directly using the synonyms list as the manner to hook to WIkipedia. For example, look at this ChemSpider Record and notice in the synonym list that three names are highlighted. Click on these to be connected directly to Wikipedia. There ARE issues with this approach...not all synonyms are valid. I have seen a post stating that there are >23000 structures on Wikipedia. Is there an easy way to find this list? Then links can be made directly. --AntonyWilliams 04:07, 29 April 2007 (UTC)[reply]

That's fantastic news! I'll help with testing. As for the 23000, that is the no. of chemistry articles on Wikipedia. I can send you that file if you like; or would you prefer us to prepare a list just of chemical compounds? Walkerma 04:27, 29 April 2007 (UTC)[reply]