Many models created in earlier versions of NetLogo also work in NetLogo 5.0. However, some models will need changes. If an old model isn't working, this section of the User Manual may be able to help you.
What issues may arise depends on what version of NetLogo the model was created with.
This guide only covers changes most likely to cause issues for users. See the Release notes for more complete details on differences between versions.
In 5.0, you don't have to put your plotting code in the Code tab anymore. Instead, you can put it inside the plots themselves, in the Interface tab.
Nonetheless, the old style and all of the existing plotting primitives are still supported. We recommend changing your model to use the new style, but if you don't, it should still work.
The following example shows how to change a model to use the new style. Suppose you have a typical NetLogo 4.1 model with one plot called "populations" and two pens called "robots" and "humans". The old code might look like:
to setup clear-all ... do-plotting end to go ... tick do-plotting end to do-plotting set-current-plot "populations" set-current-plot-pen "robots" plot count robots set-current-plot-pen "humans" plot count humans end
Here are the steps to make the transition:
plot count robotscommand and paste it into the Update Commands field for the robots pen in the plot edit dialog. Remove it from the
plot count humanscommand can be moved in the same way for the humans pen.
do-plottingprocedure, it doesn't actually do anything anymore! Remove it.
do-plottingprocedure calls in
do-plottingcall should be changed to
go, the do-plotting call should be changed to
tickwill both cause plotting to happen automatically.
The resulting (much simpler) code looks like this:
to setup clear-all ... reset-ticks end to go ... tick end
For more details on how plotting works in NetLogo 5.0, see the Plotting Sections of the Programming Guide and the Interface Guide. For details on how plotting interacts with the tick counter, read on.
The way the tick counter works has changed in 5.0. Instead of being initially set to 0, the tick counter is initially blank.
You must use reset-ticks to start the tick counter at 0 before using ticks, tick or tick-advance for the first time.
reset-ticks should go at the end of your setup procedure. Putting it there will allow your model to work with 5.0's new plotting features.
In 5.0, you don't have to put your plotting code in the Code tab anymore. Instead, you can put it inside the plots themselves, in the Interface tab. Code inside plots is triggered by reset-ticks and tick. Resetting the tick counter runs plot setup code, and then it also runs plot update code to plot the initial state of the model. The initial state of the model won't be in place until the end of setup, so that's why reset-ticks should go at the end.
In order for models from previous NetLogo versions to work in 5.0 without changes, when an old model is opened in 5.0, any occurrences of clear-all (or ca) are automatically changed to __clear-all-and-reset-ticks, which combines the effects of clear-all and reset-ticks. The two underscores on the name indicate that this is not a normal primitive, but exists only for backwards compatibility.
You should remove __clear-all-and-reset-ticks from your code, replace it with clear-all, and put reset-ticks at the end of your setup procedure. (This doesn't happen automatically because the structure of NetLogo models is too free-form for an automatic converter to reliably make the change for you.)
NetLogo 5.0 fully supports international characters cross-platform, using the Unicode character set. NetLogo 5.0 model files always represent Unicode characters using the UTF-8 encoding.
Previous versions of NetLogo allowed Unicode characters to be used in some contexts. However, model files were saved in the platform's default encoding, which on most systems was something other than UTF-8. Characters were handled correctly on the same platform (e.g. two Windows machines), but could be altered if the model was moved between platforms (e.g. from Windows to Mac or vice versa).
When opening an existing model in NetLogo 5.0, if the model contains international or other non-ASCII characters, the characters may be interpreted incorrectly, because they were originally written in a platform-specific encoding, but then read back in in UTF-8.
If only a few characters are affected, you might find it easiest just to fix them manually.
But if you expect a large number of characters to be affected, and you want them translated automatically, you can use a third party utility to re-encode your .nlogo file from its original encoding into UTF-8. After conversion, open the model in NetLogo 5.0 and all characters should be correct.
NetLogo 5.0 uses the Markdown markup language to allow you to format your Info tab, including headers, bold and italics, images, and so forth.
Earlier versions of NetLogo used a custom markup language with much more limited capabilities.
When opening a model from an older version, NetLogo 5.0 translates your old markup into Markdown. Most of the time this produces good results, but you may want to check the results yourself and make sure that your Info tab still looks good.
In NetLogo 5.0 every model has a "target frame rate" which affects the default speed at which the model runs, when the speed slider is in the middle, on the "normal speed" setting.
The default target frame rate for new models, and for models that were created in earlier versions of NetLogo, is 30 frames per second. If you are using tick-based updates, as we recommend for most models, then that translates to 30 ticks per second.
If your model runs slower in 5.0 than it ran in 4.1, it's probably just because its speed is being limited by this rate. If you want, you can press the Settings button in the Interface tab and change the frame rate to a higher number.
Some old models used the every command to set a default speed. In most case this can be now removed from the code, and the target frame rate setting used instead.
The underlying data structure for NetLogo lists has changed.
In NetLogo 4.1, a NetLogo list was represented internally as a singly linked list. Some operations on singly linked lists are fast (such as first and butfirst) but others are slow because they could require traversing the whole list (such as item and last).
In NetLogo 5.0, lists are now actually trees internally. As a result, some operations are a little slower, but other operations are drastically faster on long lists. See the Lists section of the Programming Guide for details.
Some models may run a little slower with the new data structure, especially if you make heavy use of short lists. But other models will run faster -- perhaps dramatically faster.
Some special ways of writing list-processing code that were useful in NetLogo 4.1 are no longer needed in 5.0. For example, since in 4.1 fput was fast and lput was slow, modelers sometimes built up lists in reverse order using fput, perhaps calling reverse later to restore the intended order. In NetLogo 5.0, you don't need to code that way anymore. fput and lput are the same speed.
If you are the author of an extension, you will need to recompile it against the 5.0 NetLogo.jar and lib directory for it to work with 5.0.
You may also need to be aware of the following changes:
The code for specifying the syntax of a primitive has changed slightly, for example Syntax.TYPE_STRING is now Syntax.StringType(). (From Java, the pair of parentheses at the end is required. In Scala, you can omit them.)
One significant change is that org.nlogo.api.LogoList no longer has a public constructor. Instead, there are two new ways to construct a LogoList.
If you have a java.lang.Iterable, you can copy the contents into a fresh LogoList by passing it to the static method LogoList.fromJava(). See the array extension source code for a sample usage.
Or, to build up a new list one item a time, use org.nlogo.api.LogoListBuilder. The Extensions Guide has sample code showing the use of LogoListBuilder.
In prior NetLogo versions, the extensions API required that each extension primitive have its own separate top-level class with a no-argument constructor. These limitations have now been lifted. Also, api.Primitive objects are now made only once, when the extension is loaded, instead of every time the Code tab was recompiled.
The following syntax is no longer supported:
set [<variable>] of <agent> <value>
Commands of this form must be rewritten using ask:
ask <agent> [ set <variable> <value> ]
Or, if the new value must be computed by the asking agent and not by the agent whose variable is being set:
;; OPTION #1 (using let): let new-value <value> ask <agent> [ set <variable> new-value ] ;; OPTION #2 (using myself): ask <agent> [ set <variable> [value] of myself ]
So for example, this:
set [color] of turtle 0 red
Can be rewritten as:
ask turtle 0 [ set color red ]
It is not necessary to use let or myself since red is red from the point of view of both agents.
set [color] of turtle 0 color
Must be rewritten as:
let new-color color ask turtle 0 [ set color new-color ]or
ask turtle 0 [ set color [color] of myself ]
in order not to change the meaning, since the two agents may have different starting values for color. The form using myself is briefer, but the former using let may be considered clearer, depending on context and individual preference.
Prior to NetLogo 4.0, a dead turtle's who number (stored in the who turtle variable) could be reassigned to a later newborn turtle. In NetLogo 4.0, who numbers are never reused until who numbering is reset to 0 by the clear-all or clear-turtles command. This change in behavior may break a few old models.
NetLogo 4.0 provides two different observer commands for creating turtles, create-turtles (crt) and create-ordered-turtles (cro).
crt gives the new turtles random colors and random integer headings. cro assigns colors sequentially and gives the turtles sequential equally spaced headings, with the first turtle facing north (heading of 0).
Prior to NetLogo 4.0, the crt command behaved the way cro does now. If your old model depends on the "ordered" behavior, you will need to change your code to use cro instead of crt.
It is common for old models that used crt to contain extra commands to randomize the new turtles' headings, for example rt random 360 or set heading random 360. These commands are no longer necessary when used inside crt.
Prior to NetLogo 4.0, the + (addition) operator could be used to concatenate strings and join lists. In current NetLogo, + only works on numbers. To concatenate strings, use the word primitive; to join lists together, use the sentence primitive. This language change was made to increase the speed of code that uses +.
print "There are " + count turtles + " turtles."
print (word "There are " count turtles " turtles.")
Likewise, if you need to concatenate lists, use SENTENCE.
This change is not handled automatically when converting old models; users will need to change their code by hand.
We know this change will be awkward for users who are used to the old syntax. We have made this change for efficiency and consistency. We can implement an addition operator that only adds numbers much more efficiently than one that handles several different data types. Because addition is such a common operation, NetLogo's overall speed is affected.
The observer may no longer use patch-at, turtles-at, and BREEDS-at. Use patch, turtles-on patch, and BREEDS-on patch instead. Note that patch now rounds its inputs (before it only accepted integer inputs).
NetLogo 3.1 had supports for using links to connect turtles to make networks, graphs, and geometric figures. The links were themselves turtles.
In NetLogo 4.0, instead of links being turtles, links are now an independent fourth agent type, right alongside observer, turtles, patches. The primitives involving links are no longer considered experimental; they are now fully part of the language.
Models that use the old, experimental turtle-based link primitives will need to be updated to use link agents. The differences are not huge, but hand updating is required.
Links are documented in the Links section of the Programming Guide, and in the NetLogo Dictionary entries for the link primitives. See the Networks section of the Models Library for example models that use links. There are also some link-based Code Examples.
First you will need to remove any breeds called "links" if you are only using one type of links then you will not have to use breeds at all. If you are using multiple types of links see undirected-link-breed and directed-link-breed. Commands and reporters that contain the word "links" (like __create-links-with, etc.) will automatically be converted to the new form without underscores (create-links-with). However, primitives that use a different breed name (such as "edges") will not be converted. You will need to remove the underscores by hand and unless you are declaring a link breed with that name you will need to change the breed designation to "links".
The commands remove-link(s)-with/from/to no longer exist. Instead you should ask the links in question to die.
ask turtle 0 [ __remove-links-with link-neighbors ]
ask turtle 0 [ ask my-links [ die ] ]
Several of the layout commands have slightly different inputs, the first two inputs are generally a turtle agentset and a link agentset to perform the layout on. See the dictionary entries for details. layout-spring, layout-radial layout-tutte
You may also need to rearrange the declaration of turtles-own variables, since links were once actually turtles. Any variables that apply to links should be moved into a links-own block.
Since links are no longer turtles they no longer have the built-in turtle variables (though some of the link variables are the same such as color and label. If you formerly used the location of link turtles you will now need to calculate the midpoint of the link. This is fairly simple in a non-wrapping world.
to-report link-xcor report mean [xcor] of both-ends end to-report link-ycor report mean [ycor] of both-ends endit is a little bit trickier in a wrapping world but still fairly straightforward.
to-report link-xcor let other-guy end2 let x 0 ask end1 [ hatch 1 [ face other-guy fd [distance other-guy] of myself / 2 set x xcor die ] ] report x end
and similarly for ycor.
If you used either the size or heading of the link turtles you can use the reporters link-length and link-heading instead.
We have replaced three different language constructs, -of (with hyphen), value-from, and values-from with a single of construct (no hyphen).
|color-of turtle 0||[color] of turtle 0|
|value-from turtle 0 [size * size]||[size * size] of turtle 0|
|mean values-from turtles [size]||mean [size] of turtles|
Note that when opening old models in the new version, -of, value-from, and values-from will automatically be converted to use "of" instead, but some nested uses of these constructs are too complex for the converter and must be converted by hand.
The ask command is now serial rather than concurrent. In other words, the asked agents will run one at a time. Not until one agent completely finishes the entire body of the ask does the next agent start.
Note that even the old ask was never truly concurrent; we simulated concurrent execution by interleaving execution among the agents using a turn-taking mechanism described in the NetLogo FAQ.
We have made this change because in our experience, users often wrote models that behaved in unexpected ways due to the simulated concurrency, but rarely wrote models that benefited from the simulated concurrency. Models exhibiting unexpected behavior could usually be fixed by adding the without-interruption command in the right places, but it was difficult for users to know whether that command was needed and if so, where.
In NetLogo 4.0, without-interruption is no longer necessary unless your model uses ask-concurrent (or a turtle or patch forever button containing code that depends on simulated concurrency). In most models, all uses of without-interruption can be removed.
The simulated concurrency formerly employed by "ask" is still accessible in three ways:
Note that ask itself is always serial regardless of the context in which it is used, however.
In our own Models Library, models that make use of this concurrency are rare. A prominent example, though, is Termites, which uses a concurrent turtle forever button.
NetLogo now has a built-in tick counter for representing the passage of simulated time.
You advance the counter by one using the tick command. If you need to read its value, there's a reporter called ticks. The clear-all command resets the tick counter; so does reset-ticks.
In most models the tick counter will be integer-valued, but if you want to use smaller increments of time, you can use the tick-advance command to advance the tick counter by any positive amount, including fractional amounts. Some Models Library models that use tick-advance are Vector Fields and the GasLab models.
The value of the tick counter is displayed in the toolbar at the top of the Interface tab. (You can use the Settings... button in the toolbar to hide the tick counter, or change the word "ticks" to something else.)
In the past, NetLogo always tried to update the view about 20 times a second. We're now calling that "continuous" view updates. The biggest problem with it was that you usually want updates to happen between model ticks, not in the middle of a tick, so we had a checkbox on buttons that (by default) forced a display update after every button iteration. That made sure updates happened between ticks, but it didn't get rid of the intermediate updates. You had to use no-display and display to lock them out.
We still support continuous updates. They are the default when you start up NetLogo. But most Models Library models now use tick-based updates. With tick-based updates, updates happen only when the tick counter advances. (The display command can be used to force additional updates; see below.)
The advantages of tick-based updates as we see them are as follows:
As mentioned above, most models in our Models Library now use tick-based updates.
Even for models that would normally be set to tick-based updates, it may be useful to switch to continuous updates temporarily for debugging purposes. Seeing what's going on within a tick, instead of only seeing the end result of a tick, could help with troubleshooting.
If you switch your model to use tick-based updates, you'll also need to add the tick command to your code, otherwise the view won't update. (Note that the view still always updates when a button pops up or a command entered in the command center finishes, though. So it's not like the view will just stay frozen indefinitely.)
Here are the steps to follow to convert your model to use ticks and tick-based updates in NetLogo 4.0:
Some models will require some additional changes:
Previous versions of NetLogo had a speed slider that could be used to make models run slower, so you can see what's going on.
In NetLogo 4.0, the slider can be used to speed up models as well. It does this by updating the view less frequently. Updating the view takes time, so the fewer updates, the faster the model runs.
The default position of the slider is in the center. When you're at the center, the slider says "normal speed".
As you move the slider away from the center position, the model will gradually run faster or slower.
At very high speeds, view updates become very infrequent and may be separated by several seconds. It may feel like the model is actually running slower, since the updates are so infrequent. But watch the tick counter, or other indicators such as plots, and you'll see that yes, the model really is running faster. If the infrequent updates are disconcerting, don't push the slider so far over.
When using tick-based updates, slowing the model down does not cause additional view updates. Rather, NetLogo simply pauses after each tick.
When using continuous updates, slowing the model down means view updates become more closely spaced. If you push the speed slider more than halfway to the left, the model will be running so slowly that you can watch turtles moving one at a time! This is new in NetLogo 4.0; in previous NetLogo versions, no matter how slowly you ran a model, you would never see the agents in an ask moving one at a time; all the agents in an ask always appeared to move together.
NetLogo no longer maintains an internal distinction between integers and floating point numbers. So for example:
observer> print 3 3 observer> print 3.0 3.0 observer> print 1 + 2 3 observer> print 1.5 + 1.5 3.0 observer> print 3 = 3.0 true
(The last line shows that although the distinction between integer 3 and floating point 3.0 was maintained, the two numbers were still considered equal.)
observer> print 3 3 observer> print 3.0 3 observer> print 1 + 2 3 observer> print 1.5 + 1.5 3 observer> print 3 = 3.0 true
We expect that only rare models will be negatively impacted by this change.
A benefit of this change is that NetLogo now supports a much larger range of integers. The old range was -2,147,483,648 to 2,147,483,647 (around +/- 2 billion); the new range is +/-9,007,199,254,740,992 (around +/- 9 quadrillion).
NetLogo 3.1 (and some earlier versions) included primitives called turtles-from and patches-from that were occasionally useful for building agentsets. In NetLogo 4.0, these primitives have been replaced with new primitives called turtle-set and patch-set that are much more flexible and powerful. (link-set exists as well.) See the entries for these primitives in the NetLogo Dictionary. Models that use the old turtles-from and patches-from will need to be altered by hand to use the new primitives.
In NetLogo 3.1 RGB and HSB colors could be approximated as NetLogo colors using the rgb and hsb primitives. These have been renamed to approximate-rgb and approximate-hsb and now expect inputs in the range 0-255, not 0-1.
The full RGB spectrum is now available in NetLogo so it may no longer be necessary to use these primitives at all. You can set any color variable to a three-item RGB list, with values in the 0-255 range, and get that exact color. See the Color section of the Programming Guide for details.
In previous versions __tie was provided as an experimental feature. As of NetLogo 4.0 links have a tie-mode variable which can be set to "none", "free", or "fixed". In 4.0 tie is now a link-only primitive. This means that to tie turtle 1 to turtle 0 you write:
ask turtle 0 [ create-link-to turtle 1 [ tie ] ]
See the Tie section of the programming guide for details.
If your model is behaving strangely or incorrectly, it may be because since NetLogo 3.1, agentsets are now always in random order. In prior versions of NetLogo, agentsets were always in a fixed order. If your code depended on that fixed order, then it won't work anymore. How to fix your model to work with randomized agentsets depends on the details of what your code is doing. In some situations, it is helpful to use the sort or sort-by primitives to convert an agentset (random order) into a list of agents (fixed order). See "Lists of agents" in the Lists section of the Programming Guide.
If you are seeing pieces of turtle shapes wrapping around the view edges, it's because NetLogo 3.0 allowed you to turn off such wrapping in the view without affecting the behavior of the model. Since NetLogo 3.1, if you don't want the view to wrap you must make it so the world doesn't wrap, using the new topology feature. Making this change may require other changes to your model, though. See the Topology section of the Programming Guide for a thorough discussion of how to convert your model to take advantage of this new feature.
Many models made in NetLogo 3.0 or earlier use setxy random world-width random world-height to scatter turtles randomly, using either random or random-float. It only works if world wrapping is on.
(Why? Because when wrapping is on, you can set coordinates of turtles to numbers beyond the edge of the world and NetLogo will wrap the turtle to the other side. But in worlds that don't wrap setting the x or y coordinates of a turtle to a point outside the bounds of the world causes a runtime error. The world wrap settings were added in NetLogo 3.1. See the Topology section of the Programming Guide for more information.)
To fix your model so that it works regardless of the wrapping settings, use one of these two commands instead:
setxy random-xcor random-ycor setxy random-pxcor random-pycor
The two commands are a bit different. The first command puts the turtle on a random point in the world. The second command puts the turtle on the center of a random patch. An even more concise way to put a turtle on the center of a random patch is:
move-to one-of patches