Downloading Leo - suggestions

[lewis]

I have attempted some improvements at the 'Downloading Leo' section. Goals are:
a. write a compact, focused instruction.
b. encourage Leo users to use Git
c. Keep the friendly style
d. remove distracting links, broken link (old greygreen.org), and duplication (github instructions).


Downloading Leo

We recommend using Git to download Leo. It will give you the latest full featured code. Follow the instructions at Installing Leo with git (http://leoeditor.com/installing.html#installing-leo-with-git)

Why do we recommend Git?
Leo is always being improved and developed. Leo's developers ensure that the daily commits are as bug-free as possible. You can review all the new features and update when you decide.

If you prefer a new development version, or a version from from 1, 2, 5, 10, 30 or 90 days ago, download a Nightly snapshot (http://leoeditor.com/download.html#snapshots) .

If you want a stable release download from sourceforge (https://sourceforge.net/projects/leo/files/Leo/). Remember it won't have the latest features and bug fixes though.


Snapshots
- Suggest deleting the "As of 2014" introduction. It was a long time ago and not relevant to the time frames.
[snip]


Hope it helps as you work on Leo's documentation for several more days :) These type of instructions always benefit from a review by a few people so please comment freely.

Regards
Lewis

[Edward K. Ream]

On Sat, Mar 11, 2017 at 4:59 AM, lewis <lewi...@operamail.com> wrote:

I have attempted some improvements at the 'Downloading Leo' section.

​Great stuff.  I made a few minor changes.​

 

​What do you think?

These kinds of simplifications to the first docs that newbies see are vital.  Thanks again.

Edward

[lewis]

Nice improvement. It answers the 'why git' question up front, in the introduction.

Lewis

[Edward K. Ream]

On Sat, Mar 11, 2017 at 6:17 PM, lewis <lewi...@operamail.com> wrote:

Nice improvement. It answers the 'why git' question up front, in the introduction.

​Glad you like it.

Edward

[Offray Vladimir Luna Cárdenas]

Hi,

I like how documentation is getting more compact and direct. I wonder if installation instructions could go from the most newbie friendly to the not so newbie ones (maybe after releasing 5.5). So installation should start from the most common used installation method of the intended platform and then go to more powerful ones. For example, running a binary self-contained executable on Windows, a dmg on Mac or apt-rpm on Linux. Of course that creates the problem of packaging, that we have discussed previously Leo's manpower (which is mostly Edward) could be not enough to take the laborious and tedious packaging work.

In the Grafoscopio case, I rely on the Pharo platform to manage packaging and provide prerequisites, which makes installation and updating easy, but I was wondering how Python related projects do this nowadays. One path is Conda and, particularly, Miniconda which manages depedencies and is already packaged for several platforms [1], but seems that a full installation can be overkill for Leo and its prerrequisites [2]. In the Jupyter case the map they provide [3] gives the user and overview of what they want to do, based on the answer the user gives to a question, and they use the external installer method (in their case Anaconda) to the final installation, making it pretty easy [4] ('cause the heavy work is done by conda and not the user/developer). Conda can be also used to upgrade the software, including updating/building from git[5]

[1] https://conda.io/miniconda.html
[2] https://conda.io/docs/install/quick.html
[3] https://jupyter.readthedocs.io/en/latest/projects/content-projects.html
[4] https://jupyter.readthedocs.io/en/latest/install.html
[5] http://stackoverflow.com/questions/19042389/conda-installing-upgrading-directly-from-github

I think that this is a path worthy to explore. At the end install instructions, including updating, for Leo could be (following Jupyter) something like:

~~~
# Installing Leo

1. Install miniconda for your platform.
2. Open your operative system command terminal (optional link to details on how to do it each platform).
3. Run from the terminal:
    conda install leo

# Upgrading Leo:

To update Leo to the last prebuild do open a terminal and do:

  conda update leo.

To update Leo by building it to their last development version (from git) do:

  conda build leo.

~~~

As I said, we have something similar for Grafoscopio with Pharo and installing and packaging for several platforms (Win, Mac, Linux) is transparent and updating to the last version is just two clicks away. Feedback has been positive in the workshops about this method, so some similar ideas could work for Leo.

Hope this helps,

Offray

--
You received this message because you are subscribed to the Google Groups "leo-editor" group.
To unsubscribe from this group and stop receiving emails from it, send an email to leo-editor+...@googlegroups.com.
To post to this group, send email to leo-e...@googlegroups.com.
Visit this group at https://groups.google.com/group/leo-editor.
For more options, visit https://groups.google.com/d/optout.

[Edward K. Ream]

On Mon, Mar 20, 2017 at 7:00 PM, Offray Vladimir Luna Cárdenas <off...@riseup.net> wrote:

I like how documentation is getting more compact and direct.

​I agree.  This is a big step forward, for everyone, but especially for first timers.​

 

I wonder if installation instructions could go from the most newbie friendly to the not so newbie ones (maybe after releasing 5.5). So installation should start from the most common used installation method of the intended platform and then go to more powerful ones.

​Heh.  The eternal tension. Perhaps you could have a conversation with Lewis Neal.​

​... ​

I was wondering how Python related projects do this nowadays. One path is Conda and, particularly, Miniconda which manages dependencies and is already packaged for several platforms [1], but seems that a full installation can be overkill for Leo and its prerequisites

​​

[2].

​Overkill, maybe, but I think it is useful overkill. Installing the full Anaconda package (or packages, if you install for both Python 2 and 3) saves a lot of time in the long run.  I'm not looking for anything better.  Besides, these days even small machines typically have huge memories.​

In the Jupyter case the map they provide [3] gives the user and overview of what they want to do, based on the answer the user gives to a question, and they use the external installer method (in their case Anaconda) to the final installation, making it pretty easy [4] ('cause the heavy work is done by conda and not the user/developer). Conda can be also used to upgrade the software, including updating/building from git[5]

​I think this is just fine or Leo too.

Edward

[Offray Vladimir Luna Cárdenas]

Hi,

On 21/03/17 09:11, Edward K. Ream wrote:

On Mon, Mar 20, 2017 at 7:00 PM, Offray Vladimir Luna Cárdenas <off...@riseup.net> wrote:

I like how documentation is getting more compact and direct.

​I agree.  This is a big step forward, for everyone, but especially for first timers.​

 

I wonder if installation instructions could go from the most newbie friendly to the not so newbie ones (maybe after releasing 5.5). So installation should start from the most common used installation method of the intended platform and then go to more powerful ones.

​Heh.  The eternal tension. Perhaps you could have a conversation with Lewis Neal.​

If the miniconda method progress, Lewis could help in creating the yaml manifest for Mac, instead of the brew installer... I started thinking in this, installation method, but as the mail progressed miniconda seems

​... ​

I was wondering how Python related projects do this nowadays. One path is Conda and, particularly, Miniconda which manages dependencies and is already packaged for several platforms [1], but seems that a full installation can be overkill for Leo and its prerequisites

​​

[2].

​Overkill, maybe, but I think it is useful overkill. Installing the full Anaconda package (or packages, if you install for both Python 2 and 3) saves a lot of time in the long run.  I'm not looking for anything better.  Besides, these days even small machines typically have huge memories.​

Yes. I will start with miniconda and see what we can bootstrap from there. Connectivity is not the same in the Global South and Leo has users in India and Colombia. Reducing the amount of stuff you have to download to start with Leo is important to make it global.

In the Jupyter case the map they provide [3] gives the user and overview of what they want to do, based on the answer the user gives to a question, and they use the external installer method (in their case Anaconda) to the final installation, making it pretty easy [4] ('cause the heavy work is done by conda and not the user/developer). Conda can be also used to upgrade the software, including updating/building from git[5]

​I think this is just fine or Leo too.

I agree. I will try miniconda for my python related stuff (is not much these days) and try to give feedback.

Cheers,

Offray

[Offray Vladimir Luna Cárdenas]

On 21/03/17 09:48, Offray Vladimir Luna Cárdenas wrote:
>
> If the miniconda method progress, Lewis could help in creating the
> yaml manifest for Mac, instead of the brew installer... I started
> thinking in this, installation method, but as the mail progressed
> miniconda seems
>

Whoops! hit enter at the wrong time. I meant that I started my mail
thinking in native installers, but as I progressed the (mini)conda
method seemed better.

Cheers,

Offray

[Edward K. Ream]

On Tue, Mar 21, 2017 at 10:06 AM, Offray Vladimir Luna Cárdenas <off...@riseup.net> wrote:

Whoops! hit enter at the wrong time. I meant that I started my mail thinking in native installers, but as I progressed the (mini)conda method seemed better.

​Good.  I think we are in reasonable agreement.

Edward

[lewis]

Ah yes, feel the eternal tension :)

Miniconda or Anaconda? 

Using Anaconda, you have a very large set of packages, some of which are outdated.

From the Conda.io https://conda.io/docs/install/quick.html site:

NOTE: If you choose to install the full Anaconda package, it requires 3 GB of available disk space.

Using Miniconda  py3.6 64bit the Binary File is only 57.8 MB.

For me I would probably go for Miniconda as it allows more control over packages, although I tried it only once and it didn't play well with packages I already had installed.

At first glance my preference might be to have a separate section for those interested in installing Ana/Mini conda, rather than complicate the 'compact and direct' docs.

Edward - do you have any statistics on how Leo's user base installs Leo? Sure the numbers might be meaningless as most git downloads are updates, but the knowledge may help guide the direction of the install documents.

HTH however I don't have a Mac. 

Regards

Lewis

[Edward K. Ream]

On Tue, Mar 21, 2017 at 9:59 PM, lewis <lewi...@operamail.com> wrote:

From the Conda.io https://conda.io/docs/install/quick.html site:

NOTE: If you choose to install the full Anaconda package, it requires 3 GB of available disk space.

Using Miniconda  py3.6 64bit the Binary File is only 57.8 MB.

​Thanks for this.  This might be significant for some people​, but I wonder how many.

Edward - do you have any statistics on how Leo's user base installs Leo? Sure the numbers might be meaningless as most git downloads are updates, but the knowledge may help guide the direction of the install documents.

​There are stats on Leo's download page on SourceForge.​

 

​

Edward

[Offray Vladimir Luna Cárdenas]

For me and in this context certainly 60 MB versys 3 GB will be significant. I think that we should not only worry for the current users but about future ones in several places with low connectivity. I would try miniconda first to see how Leo and its dependencies can be packaged there. If that work the same recipes will work with full and big Anaconda installs, but 3 GB is not a prerrequisite to start with Leo, just a possibility.

Cheers,

Offray