- Now namespace is normalized when new project was added to the service.
- User profiles were added. Now you can view user's avatar, which projects he is moderating or tracking. To change your avatar, you need to setup an email at account settings and to upload the avatar to gravatar.com.
- Email field is required now.
- And on each project's page you can see who tracks and moderates it.
- Project documentation is reformatted into reStructured text because of it's flexibility.
- Now versions with release date "in future" are concidered as "unreleased".
- If there are words "in progress" somewhere in version description, it is marked as "unreleased" now.
- Fixed an error which prevented changelog building from commit messages for some repositories.
- Fixed issue with reporting problems about projects with non ASCII names.
Now user tags are shown in the email digests, Slack and Webhook notifications.
Algorithm for collecting release notes from Git's commit messages was fixed and improved. Previously it produced wrong results in some circumstances.
- Now links to tagged project contains anchor name to open right tagged version.
We are constantly working on improving AllMyChanges' usability. The one thing which was really missing in our service is ability to mark some versions as "used in that project". Services like VersionEye or Gemnasium bind everything to what they call "projects" – first you create a project then push it's requirements and track updates.
But in AllMyChanges we went another way – we let you to track any release notes and to tag them as you wish. "Tag" can designate a project you are working on, an environment, or a hostname. In other words, "tag" is everything what is makes sence for you use cases.
First of all, tags are remainders which helps you to remember which version you are using. This also helps you to decide if you need to upgrade to a newer version.
To make it more convenient, we've added two new pages. One shows all tags you've created and number of new versions in all release notes tagged by these tags. This page is useful to overview all you project or environments and see how outdated they are.
Other page let you go deeper and investigate details about release notes, tagged by single tag. This way you will see this libraries are outdated and require more attention. And of cause, you can go deeper and dive into their release notes to check if you need to upgrade right now or there are some backward incompatibilities.
Project versions can be tagged in three ways:
First and simpliest – on the web, by pushing "Tag" button besides version number in release notes.
Second – using command line client amch. This is as easy as:
amch tag django 1.8.2 allmychanges.com
Here we tag 1.8.2 version of django with tag allmychanges.com.
And finally, third and recommended way is to build automated
syncronization process which uses amch command line tool to
upload used versions and to tag them in batch mode.
Command amch can be used to import data from csv format, which
should look like that:
namespace,name,version,tag
python,alabaster,0.7.7,allmychanges.com
python,anyjson,0.3.3,allmychanges.com
python,argparse,1.4.0,allmychanges.com
python,args,0.1.0,allmychanges.com
To create such csv from Python's requirements.txt file, we've
created a simple utility pip2amch, which can be used like that:
pip2amch --tag my-project requirements.txt | amch push
Here pip2amch will take a requirements.txt file and output
to stdout csv data in a format required by amch command.
Next, amch pushed data to the AllMyChanges and tags all versions,
specified in input data.
This way, you'll have these versions fixed in allmychanges. If you are
working in other ecosystem, you can easily (or not, depending on your
package manager) write your own requirements parser and use it instead
of pip2amch. This is how AllMyChanges differs from Gemnasium and
VersionEye – they parse requirements only for a limited number of platforms
and make you deal only with projects having those requirements.
As some API handles were changed, it is recommended to update amch
command line client to the latest version. If you are a python developer
install pip2amch and automate requirements pushing to AllMyChanges,
otherwise, write a tool, similar to pip2amch and again, automate :)
Use tags to check is one of your working environments have outdated requirements or libraries.
Now it is possible to add changelogs without name or source url, but their tuning should be finished before they will be crawled by robot.
Also, on every project's page, you'll see a warning message if you have some projects which are not tuned. Following a link in the message, you'll see a full list of projects which should be finalized.
- Now we return
changelogfield for tags. - It is possible to filter versions by changelog's namespace, name and version number.
- Now changelogs can be filtered by keyword
id__in. - Handle
/user/was added to check if you are authenticated or not. In case if not, it returns 401 HTTP status code.
- Fixed usage of symbols like
#in namespace or project's name. Thanks to @hawkunsh for report about this problem. - Fixed guessing html page encoding when there is Content-Type header without charset, and meta tag in html page with charset.
- Fixed html to markdown processing. It shouldn't break long links now. This fixes some issues in Slack notifications where some links appeared broken.
- Sticky version plate was made more predictable. All known bugs were fixed.
Some parser improvements were made to make it possible to parse Zabbix's release notes:
- Now "Tune" page reports to the user which HTML pages were downloaded by recursive HTTP downloader.
- Now all
.phpfiles considered as havinghtmlmarkup. - Now if there is a "search list" given by user, then files discovery mechanism does not takes place.
- Fixed regression – text "NO FALLBACK DEFINED" in Slack's notifications.
- Now Slack notifications are sent as attachments and large release notes are hidden under the "Show more..." link:
- Allowed whitespaces in a link description when preparing notification for Slack.
- Now images are presented as urls when posting notification to Slack.
- It is possible to escaped slashes in
sedchangelog preprocessor's rules. This way, some items in the text can be replaced with urls. For example, this feature was used to link GitHub's and Launchpad's issue numbers to their actual pages for python/lxml.
- Fixed urls and bold text representation in Slack notifications:
- Fixed issue with updating projects where release dates are known.
- Fixed appstore and google play release notes headers.
-
Algorithm which adds versions to the digest was significantly improved. Now it shouldn't add many old versions for a new or unpaused project.
-
Now Slack and Webhook integrations can be tested by pressing "Test" button on the account's settings page.
-
Also now we post full text of release notes to the Slack.
-
Now we more precisely describe at account's settings page, how often email notifications will be sent. Previously, some users were confused because switching digests to 'Never' turns off email notification completely and this wasn't obvious.
The main feature of this release, is a new way of collecting changes for email digests. Previously there was an issue with these digests. Some project maintainers keep a section with unreleased version's description in the changelog. Old algorithm discovered these versions but didn't sent them in an email when this version has been released.
New algorithm, discovers the setuation, when known "unreleased" version becomes "released" and includes it into a next digest for every user who tracks the project.
Added support for one number versions, but these
should be prefixed with v or r, like v15 and r2015.01.02.
This change made possible to parse release notes for projects which use simple numbers for versioning. Here are few examples:
- Fixed error when comparing versions like 0.1.2-rc3.
- Allowed
supandsmallhtml tags in changelog markup.
- Fixed another place where snapshots are generated.
- Page with "Digest" was deprecated, now you'll see a link to the page with a list of your tracked projects.
- Fixed parser's super power to build changelog from git's commits. This is a huge improvement because some time ago, this feature was broken and we were unable to collect changelogs for some projects. Now we are back in business :)
- Fixed
=====in Google Play's release notes by using oneline format for markdown headers. - Fixed our tool which makes a changelog's screenshot each time you share an URL to twitter or facebook.
- Changed placeholder's text in the magic prompt.
- Added ability to view projects by id, using urls like
/p/54/. But this is more for convenience during log viewing and troubleshooting sessions. - Added changelogapp (nodejs command line utility) to a changelog generators list.
We finally migrated to dockerized deployment. This opens following possibilities:
- it will be easier to scale in case if somebody suddenly will share a link at https://www.producthunt.com;
- it will be easier to create "enterprise" version of AllMyChanges which will be hosted by big companies in their intranets.
- it will be easier to setup development environment for a new team member. We are not hiring! But you can join the team if you want to build something useful for community.
In case, if you have any questions, write me to mailto:[email protected].
New mechanism of package tuning and downloader selection. Massive changes were made not only on the frontend, but also at the backend.
Now parser is able to process changelogs which are separated to many files and where versions are in the filenames, like in openssh's release notes. Here is the result of parsing these release notes.
- Some face lifting was made. Thanks to @vovanbo for css diff!
- Use 'updated' date of atom feed if there is no 'published' date. This allowed us to process Spring Boot's and Spring Framework's changelogs.
- RSS integraion's field is 'readonly' now instead of 'disabled'. Thanks to @Splurov.
- Fixed Google Play's URLs processing when there are parameters other than 'id'. Thank again to @Splurov.
This release includes several important improvements:
- Better google play integration. Now it works even for apps which have different versions for different devices.
- Support for RSS/Atom feeds as a datasources was added. This allows to parse blogs' feed and now we are able to process Ruby's and Dropbox's changelogs correctly.
- Javascript version of our packages catalogue was replaced with oldschool one. And now it show be more usable. Go and check it yourself.
By the way, if you were confused by "Add New" page's complexity, than I have good news – next release will include a redesign of this page. We are continually improve the service thanks to your reports. Don't forget to sending us yor suggestions via email or gitter chat.
We've added a new channel for reading your favorite release notes. Now you can subscribe add a feed into your favorite RSS reader.
To get you personal feed's URL, go to the account settings page and discover it in the "integrations" section.
Thanks to Denis Gladkikh for pushing us into this direction.
First of all, I want to say "thank you" to all russian developers who came from habrahabr to test our service. Thank you guys, you've helped to find many issues and to improve quality of the service.
Because of you reports we now, for example, we now
parse OpenSSL's release notes correctly.
To do this we have had to add a special preprocessing
engine which is like sed is able to prepare texts
before our robot will try to find there release notes.
- We've fixed errors which prevented to parse FreeBSD's release notes.
- Fixed an error in git commit messages parsing which prevented processing of Laravel Framework
- Some other minor tweaks and fixes.
Again, thank you guys! Keep adding new projects and reporting on issues!
Did I say we implemented a generic web-hooks? Because we really did. And now you can be not only receive notifications in Slack, but do anything you want when a new version of a package is found. Read more about web-hooks in our documentation.
To help you to automate your processes, we've prepared an example script which uses a python-processor to accept a web-hook and to post notification further into different Slack channels and email, depending on package's namespace.
- Also, now we or processor is fixing header levels in GitHub Releases by replace
all
h1headers withh2headers, allh2withh3and so on. - Now package name is added as a hashtag when our robot @NewReleaseNotes tweets about a new release.
- Annoying popups with tips wer disabled while we are implementing a mechanism to remember which tips you already seen.
- We started experimenting with cal-heatmap javascript plugin to show you your activity on the service. But it is not ready yet.
Many of you were begging us to make AllMyChanges work with GitHub's releases API. And now this feature was implemented.
Among 700 packages in our database, only 500 are hosted at the GitHub and 100 of them are using GitHub's Releases. Now for all of these projects we can provide even better release notes.
Here is piece of libsass's release notes collected from git commit messages:
It is amazing how hand-written release notes are better:
And GitHub Releases are available for 20% of projects, hosted at the GitHub now.
Subscribe on them at AllMyChanges and we'll deliver new release right into you mail box.
- Now we allow embedded video players in release notes.
<br>tags were fixed in sanitizer.- Screenshots published into the twitter @NewReleaseNotes were improved.
This release includes two major features: "track list" page and "sharing like a pro". First one will let you checkout which packages you are followed and sencond one will make easier to share new releases to you friends on twitter.
A lot of users were asking this feature. They wanted to see a list of packages they are following on one page. Now this is possible.
Open this page by clicking this item in a dropdown menu:
In a list you will see packages with their descriptions and latest version numbers. All sorted alphabetically. If you have any ideas how to improve this page, please tweet to @allmychanges.
Looking how often people are sharing pieces of release notes as a screenshots, we decided to implement Twitter Cards specification to show parts of a changelog as an image right in a twitter timeline.
Sadly, but this does not work. Twitter hides images from twitter cards by default and you have to click on a tweet. So, we started a small research to find out a best way to share fresh release notes with your friends.
And solution was found! Now, when our robot discovers a new version of a package, she makes it's screenshot and posts it as a tweet to a @NewReleaseNotes account.
For versions which were twitted this way, we show standart "Tweet" button on web and email:
Clicking on this button will open a new window where you can retweet the original post:
And you friends will see this retweet with a screenshot of the changelog:
If you don't like an idea to retweet, then just share a link to the changelog and twitter will show a twitter card for it.
Hey, are you still reading this? Go and test these new features!
- Login via GitHub button was fixed at popup appearing after clicking on "Follow" button. Closes issue #30. Thanks to Denis Gladkikh (@outcoldman) for reporting.
- Now tooltips are shown only after 15 seconds of inactivity at the page. More improvements are waiting in our roadmap.
Them main feature of this release is support for Google Play URLs. Find your favorite Android apps on Google Play, paste their URLs right into the AllMyChanges's search bar and subscribe on the release notes.
Secondary change we would like to introduce is a better and smarter changelog suggestion on the front page. Now it takes into account which packages did you track or skip in the past. There isn't machine learning behind yet, but now it is much more interesting to flip through different packages, viewing their latest release notes. Did I say you shoudl try it? No? Okay, Go And Try It!
- Now we have a FAQ page.
- Also, another help page was added – Changelog Generators. It should help you to keep better release notes spending less time writing them.
- Few minor changes were introduced, like renaming of "Track" button into "Follow" button, etc..
- Also site now uses gorgeous Intro.js to show tooltips when you hang on a page for more than 15 seconds. We are thinking how to make this feature less annoying and more helpful. If you have any ideas, please share them with us.
This is release of minor improvements.
- Whole site went behind a HTTPS.
- Typography was fixed here and there.
- Buttons become more colorful and gradientish.
- Functions
amch:re.subandamch:re.matchwere available in XSLT transformations. - Now we show spinner while waiting for autocomplete, because it is not optimized yet and sometimes takes really long to go through a billion of iOS apps, searching a term.
- Recursive http downloader was fixed to not add .html extensions to files with other content type.
- When anonymous track a package we show him (or her) a popup, explaining that to receive notifications he (or she) should login.
- And finally, now we show action buttons (report, track, etc.) on a sticky header along with version numbers.
This release includes two major features:
-
Design became responsive and more usable on devices with a small screen. Though not all pages are fixed yet, but frontpage and changelog os single package are looking good now on my iphone :)
-
Now XSL transformation could be applied to html pages before our robot will make attempt to find version descriptions there. This allows to fix weird HTML markup which brokes our data extraction algorithms.
This powerful feature is accessable only to service creators. It allowed us to process release notes of Minecraft, 3D Engine Unity and Chrome. We hope to add more in the future.
If you encounter a changelog which looks good on the web, but service is unable to handle it, email us, and we'll be happy to help.
Also, there were a few minor changes:
- A bunch of SEO optimizations were introduced like micro formats support, meta tags, last-modified caching, etc..
- Fixed date parsing for dates ended with dot, like "Released on Dec. 5, 2014."
- Added description of
amchcommand line utility to the tools page.
New autocomplete system was introduced. Previously it worked only by our own database, but now, when you are entering something into the search bar, it also searches through over 1 million apps from the Apple AppStore.
Very soon we'll add other sources for autocomplete, such like PyPi, CPAN, NPM and GitHub.
From today, each package on AllMyChanges have new attribute "description". It is optional, but we highly recomment to fill it. For github projects we filled it automatically and are planning to do this when new packages are added.
Now when tuning how changelog is parsed, you can enter special filters into the "search list" textarea:
[title=Release notes for \d.*]
This especially useful with rechttp downloader which
downloads bunch os html pages and tries to figure out which
parts of them are version descriptions. Using such
"attribute filters" you can help the parser to understand
which html headers are really contain version numbers.
- Rechttp downloader now adds '.html' extension to urls without backslash at the end.
- Fixed error on a page for receiving a token when user come without login.
- Now we show number of trackers beside the "Track" button.
- Fixed a way how version numbers are compared when we are sorting them (distutils's LooseVersion uses wrong algorithm).
- Fixed version parsing when there are upper-case letters in a version number 5.3.BETA-22.
- Don't send Slack notification when no new versions were found.
This release bring significant improvements in our notification system. Now you could go into your account's settings and to select if you want to receive email digests daily or weekly.
Or, you could turn email notifications off and enable Slack notification instead.
- Go to you Slack account.
- In channel's dropdown menu, choose "Add a service integration..."
- Next, filter all available integrations by "webhook" keyword.
- And select "Incoming WebHooks".
- Choose a channel to which notification should be sent by AllMyChanges.
- And push this huge "Add Incoming Webhook Integration" button :)
- Finally, copy value from "Webhook URL" field into "Slack Url field at allmychanges's account settings page.
Also, our downloader was improved and now it is able to fetch multiple pages. This is required some tuning, but works pretty good. For example, whithout it we weren't able to process release notes of some important projects, but here they are:
Take look at the settings of these projects (push "Tune" button on their pages)
to understand how rechttp downloader works.
-
Fix version parsing so that versions like
2.0-beta.6.7,40.0.2214.73,v.1.0.3.2and2.0.1rc1now parsed correctly. This made possible to add new release notes of Textmate, Chrome for iOS and suricata. -
Nicer formatting for lines like:
Bug #1184: pfring: cppcheck warnings Feature #1309: Lua support for Stats outputparsed by plaintext parser.
-
Also, plaintext parser now used
<br>for textwrapping within paragraphs. -
Now we use
utf-8as default encoding for files downloaded usingrequests. This fixes issues with unicode when server don't return encoding in the http headers.
This release brings an amazing feature. Now you can track a changelog of any iOS application!
It is very very easy. If you didn't find the app
in our catalogue under ios namespace, then
search it in the AppStore and right click on app's
icon to open context menu. It could look like that:
Or you could open drop down menu inder the apps icon:
Select "Copy Link" menu item. Open AllMyChanges.com and paste app's url into the search input at the top of the page.
After hitting the "Search" button you will see as service will process release notes and show it as usual. And good news are -- it is trackable, and you even don't have to install the app on your device!
P.S. -- if you are wondering if only latest Slack's release note was so verbose and nice, go and discover their previous release notes. I wish all app developers were like slack's developers not like facebook's :)))
- Plain text parser was fixed that way so now it is able to parse nodejs's changelog correctly. By the way, if you are interested in server-side javascript, then you could also track iojs's changelog too.
- Source guesser was optimized to extract urls not only from html attributes but also from many other weird places. And, now it sorts results so that better matching come first.
- Also, guesser for python was slightly improved and now is able to find right repository urls for packages like django-extensions.
- Fixed issue with Django's 1.7.6 version which was parsed as released, but actually is "under development".
- Fixed error with unicode filenames in repository.
- URL validator was fixed to accept urls like
git@github...andhg+http://bitbucket....
The main feature of this release is a sticky navigation bar for package's version.
Previously it was very hard to navigate through long changelogs because large number of versions or because their description was very long. Now this problem is solved by special navigation bar which sticks to the top of the screen and shows you a version number you are looking at and also all other versions. And you can click on any version to open it's description.
Other changes:
- Now service generated SVG badges instead of PNG. And they should look nice on retina displays.
- Keyword 'bugfixes' now highlighted.
- Now links to packages and their versions are blue in emailed digests.
And bugfixes (any software has bugs, right? :)):
- Fixed issue when version has a date but considered "unreleased" because a keyword in description.
- Fixed link from logo text to index page.
This release introduces significant changes in a way how AllMyChanges stores and shows information about package's versions.
Previously, it tried to figure out where a list items in version description, and store them separately on database level. Also, our algorithm tagged these items with labels like "fix", "new" "sec" and "dep". This approach produced nice results in very simple cases and awful in relatively complex. For example it was unable to handle nested lists and somitimes didn't work with complex html changelogs.
Today we changed our processing pipeline, and will store information about each version as whole text. Then it is processed and some keywords are underlined to attract you attention to bugfixes and security issues. We hope this algorithm will work better with complex changelogs and will provide better user experience.
Right now not all packages have migrated to this new scheme, but all of them will be converted sooner or later.
As usual, if you have any ideas how to improve the service, feel free to drop a line via our support email or through a feedback form available via click by "Report" button on package's page.
This version includes many bugfixes and improvements. The most significant are: header's redesign, packages catalogue and step by step guide after the first login.
Now header includes a tree links to the main services functions. Using these links, you could "add a new package", "navigate through existing packages" and "view your digest".
Also, search input was completely rewrote in react.js and now should work on every page even on the angular's. Now search results can contain three types of results: namespaces, packages and urls found for given namespace and package's name on pypi. In latter case, you could select this url to add it as a new package.
In other words, search become much more convinient now. Just go and try it yorself!
Packages catalogue is a browsable directory of namespaces and packages known to the service. This way you could discover interesting packages and track their changelogs.
This is very important change because previously allmychanges was too odd for a newcomer. It didn't tell him what to do after the registration.
Now service tries to guide user through setup and describes how to search changelogs on the site and how to add new packages.
- Now VCS parser is able to extract version information from git tags!
- Parser removes 'version bumps' from version descriptions extracted from VCS.
- When extracting changelog items from VCS history, newlines are replaced with html
<br>tag. - Version ordering at package's page was changed to more natural however sometimes still wrong.
- Now we use HTMLParser with 'utf-8' encoding when parsing html, this fixes processing unicode symbols in markdown changelogs.
- Now git history extractor processes branches properly.
- Fixed an error in VCS parser when last commit is a version bump.
This version contains many internal improvements and should improve parsing quality. VCS changelog extractor was significantly improved. Now it produces more correct results and also it became 5 times faster. Few error in python version extractor were fixed and now we should better group commit messages for python projects. For example, previously we did very poor job for SleekXMPP, but now are able to show all latest versions correctly.
These changes should fix the amch, command line client to AllMyChanges.
- Now all source urls are normalized by API handles before changelog creation, update or search.
- Now changelogs could be filtered by source in API like
/v1/changelogs/?source=.... - Also, now API accepts urls starting with
git://....
- Better filtering of versions when there is version number in a filename and it is absent in the content.
- Fixed issue when parser ate some parts of text.
- Fixed way how changelogs are parsed when they are in a separate files where filenames include version number like in case of Go.
- Now algorithm ignores lines with more than 5 words when searching version numbers.
- Version description's typography was significantly improved. Now you can read Django's changelog and it won't hurt your eyes.
- Now preview's results are copied to an added package's changelog. This improves user experience, and seems more natural because you haven't to wait anymore.
An issue reporting was added for packages. Use new button "Report" to open popup and fill the issue. You will find this button near a package's title.
Please, write about any bugs, problems and inconsistencies, found on site. This could be missing or wrong versions, bugs in a text or date parsing, etc.
Right now this feature is in it's alfa stage. There is no ability to communicate or even view issues you've filed. But we are working on improving this feedback tool and will add comments and email notifications in near future.
Leave your feedback, it is highly valuable for us.
- Fixed date parsing for cases like in Synology's Disk Station's log, where version numbers like 4.2-3252 were parsed as dates.
- Fixed date parsing when date contains abbreviated month ending with dot like
Aug. 17, 2012. - Version parser's error was fixed. Previously it parsed
v3.0.0-preas3.0.0. - Strip newlines and spaces from file section's titles when parsing html changelog. This fixes lodash's changelog parsing.
- Fixed graphic badge generation for version with minus sign in them like
2.3-pre. - Now links to github's wiki or direct links to a raw files could be used as a changelog's source.
- Fixed issue when newly added packages were not updated regularly.
- Fixed error when file with .md extension could be considered having a reST format. This caused wrong changelog parsing for Ansible
- Improved discovery of unreleased versions. Now we have a special checker which compares if latest version hasn't a release date but few previous have. In this case, version without a date considered unreleased. This could be checked on the same project — Ansible. It's maintainers add new feature and bugs description in a changelog and when they decide to ship a new version, they just specify a release date. Currently they are working on 1.8 version.
- Now VCS history extractor uses a new pipelined workflow and items' labels were fixed.
- Critical error was fixed, when all changelog's version were deleted before each update scheduling.
This release includes big refactoring of changelog and preview update process. It is almost invisible for the end users, except now you will see separate preview stages during preview generation. I have ideas how to improve it even more, but will leave this for a while.
Also, this release includes these minor changes:
- reST format recognizer was improved.
- And now reST parser removes field-list tables, so, for example, Celery's changelog looks much nicer now.
- Now changelog items's type labels are embedded into the HTML when changelog is parsed. This should improve changelog's rendering when items contains complex markup like paragraphs in the lists.
- Changelog styles was changed a little, to fix inner lists' margins.
- Also, search list saving on 'tune' page was fixed.
- Was fixed html changelogs type guesser and now sbcl's changelog is parsed correctly.
- Fixed changelog items classification for sbcl and haproxy.
- Our parser now works better with freak dates like
September 1st, 2014and versions likev2.0.0-beta.1. This change will allow us to track javascript/handlebars, javascript/meteor and some other projects. - Source type guesser was slightly improved and now is able to understand that GitHub Enterprise's changelog should be downloaded via HTTP, not via Git.
- Also, html changelog parsing was significantly improved and now abovementioned GitHub Entreprise's log looks relatively well.
Now downloader support branches for git repositories.
For example, Celery uses separate branch to maintain
current stable version 3.1.x. And to gather
correct changelog, you have to specify correct repository
url: https://github.com/celery/[email protected]. Pay attention
to the @3.1 suffix. This way you specify a branch to be
checked out before our crawler will search and process
changelog.
Moreover, in this release:
- crawler become smarter when parsing documents in reST format.
- package view was fixed to not show versions created during preview.
- more grammatic fixes by Yuri Artemenko.
- also, there is an internal improvement which allows us to track is our changelog parser was broken after latest release.
- New account creation issue was fixed (thanks to Vasily Shmelev for report).
- A bunch of grammar mistakes were fixed on the front page (thanks to Yury Artemenko).
Today we released a new feauture. Now service's main page don't redirect to http://allmychanges.com/digest/ anymore but shows 20 most recently updated packages along with their versions.
That way it is really easy to track new packages.
- Fixed error when old versions appeared in the digest.
- Magic prompt's "Search" button was fixed.
- Now source downloader is selected automatically.
URL prefixes like
http+orhg+are obsolete now.
At the beginning of the project, we've had a separate workflow when you come across interesting project at the GitHub, click a bookmarklet and allmychanges.com searches and displays project's changelog.
This behaviour was broken because of refactoring, but now it come back.
Go to the /tools/ page to install new bookmarklet!
This is a big release which completely changes a way how packages
are tracked. Previously, you have to open "Edit Digest" page and
to add packages one by one to the list. This interface
was completely replaced by the new one. Now each page contains
a magic search prompt. Try to enter there a package repository's
URL or just "namespace name" pair. The latter works only for python
and perl namespaces but will be extended to npm and ruby gems soon.
If allmychanges.com know nothing about entered source url, it will offer you to add it and track. At the same time, it will try to fetch changelog data and display them as soon as possible, allowing to tune some crawler's parameters.
Another big change is new changelog parser which now works for everyone by default. It is still immature, but we'll do our best to improve it's quality. If you'll discover any errors or have ideas, please, write us an email.
- Improve new parser's quality.
- Simplify entering package description by automatic data gathering.
- Add an index page for logged in users. It will suggest packages to track.
Try new features and stay tuned!
Only two visible changes are introduced in this version.
- Digest editing page was renamed to 'Edit track list' and some links were made buttons.
- There was made a minor improvement in the way how changelog labels are rendered. This should fix they rendering in some mail clients, if don't, then contact us via [email protected].
Another big feature which will come soon - global magic pony search. Stay tuned!
If you are writing changelogs for your own projects, please keep dates simple and unambiguous. Something like YYYY-MM-DD fill be ok. Teach your parents too ;-)
- Now for dates like 10.08.2014 parser takes first number as a day. This should fix issue with django-taggit's changelog.
- Colored changelog item's labels in emails were fixed. Now emails should be much more readable. Thanks to @bessarabov for a report.
- Now all package urls are normalized upon save. This is most visible
for GitHub hosted projects. For example, if you are submitting
a url like
[email protected]:sass/sass.git, it will be transformed tohttps://github.com/sass/sass. - Now new parser only checks files if they have proper extension or it's filename looks like changelog.
- Fixed issue when new parser extracted version number from script html tags.
Now parser is able to group changelog items, extracted from VCS history of Node.js packages. Horay! This means that we could build a changelog for nodejs packages even if there is no handwritten changelog!
There are some other changes too.
- Fixed parsing of dates like 'August 13th 2014'.
- Now VCS history parser ignores empty commit messages.
- Scope where parser searches unreleased keywords was limited to first 300 symbols of version's description.
- Don't consider version number in case if it is preceeded by some symbols like backslash and other punctuation. Fixes cases with HTTP/1.1 parsed as a version number.
- Added ability to specify files and directories where to look for changelog data. These hints could help parser to recognize data in a vague situations.
- Now parser sanitizes HTML in his input files and escapes dangerous tags.
- Fixed issue when job queue was filled with numerous identical task which led to long intervals between package updates.
- Job's timeout was increased from 3 minutes to 10. This should help to process big repositories or extract logs from long VCS histories.
Both these changes are related to the new parser which is not available for wide audience. If you wish, create an issue and we'll add you to the experiment.
- Added parser for plaintext files which are not looks like Markdown or Rst. This significantly improved perl packages' changelogs parsing.
- Fixed way how bullet changelogs are parsed (previously parser just ignored them). Now python-redis's CHANGELOG is parsed correctly.
- "Today" section of the digest was renamed to "Last 24 hours" because it is better describes an inner changelog discovery mechanics.
- Now service uses new fonts from TypeKit. More thin and clear. Please test it and send us your suggestions.
- Fixed issue with perl/Mojolicious package where maintainer adds an empty unreleased versions into the changelog.
- Fixed error when parsing multiline changelog items. There was an issue with embedded newline characters and the way how these items were renderer. This is related to the new parser only.
- Fixed error during nodejs's changelog parsing, however it is still
- Now parser is able to process dates like 2014/05/17.
- Also it've got an extended support for month abbreviations in parsed dates.
- Fixed way how dates parsed for perl's boolean package.
- Now we think version is not release if there is 'not yet released' keyword in it'. I think we should put online documentation with description of all keywords and logic behind this mechanism.
- Now all code snippets for badges have a utm_source argument. Please, update you badges.
- Links to badge and package versions were fixed.
This release includes a new changelogs processing pipeline, which now runs in parallel to the old parser. We are keep eye on it's quality and performance and will switch to it in the near future.
Also, there were some other changes:
- We've fixed how source suggest works for perl packages with two colons in name.
- OAuth token's field was made wider and now whole token is visible.
- Now our robot is able to recognize CVE security issues and to mark changelog's items as security, deprecation or backward incompatible. Look at the Nginx's and pip's changelogs to understand how it works.
- Also, missing values in 'Last update was' and 'Time to next update' were fixed.
Few cool feature, related to usability were added recently:
- Now service is able to guess source urls for packages in
perlnamespace. - And we've implemented OAuth provider, allowing to write a third party tools and integrations.
Right now there is a simple interface to receive an OAuth token
and only one tool, implemented by myself — command-line client,
which have import, export and add commands.
Also other changes were introduced:
- Cleaner design (thanks to Maxim Sukharev a.k.a @smacker).
- A number of bugs were fixed.
More than month passed from the previous release and we'd like to present a newer and better version of AllMyChanges.com. A number of bugs were fixed and many improvements were made. Actually, most of them were already tested in production because we are using continuous delivery.
Here are most visible improvements, grouped by topics.
- Nice badge generator now shown on each package's page.
It looks like this:
and could be pasted into theREADME.mdalong with other badges from Travis and PyPi. This way viewers of you repository will be able to examine all changes in one glance without searching changelog in deeps of a source tree.
- Now our robot is able to parse dates like this "May 21st 2014".
- Also, we now do our best in parsing changelogs with nonstandart formatting, like Nginx's log. It was really hard to produce this nicely formatted changelog from it.
- Version extraction for python packages was significantly improved.
It is used when there is no human written changelog. In this case, our smart robot
goes through VCS history and tries to reconstruct changelog from commit messages.
Previously it worked only for python packages which use
setuptoolsin theirsetup.py, but now we are using cool hack and version extraction works even for packages withdistutils. - Unreleased versions could be detected if there is a word
unreleasedin it's title. Here is an example:Version 1.0.0 (unreleased). We'll understand that this version will be released in future and won't show it in the digest.
- All new users have on default package in their digest -
web/allmychangesas an example. And this package was added to the digest of all old users. If you want to receive news from the AllMyChanges.com, just leave this package in your digest. - A special page was made where you is able to setup email and timezone. Please enter correct email to receive digest propertly.
- All digest email are sent around 9AM according to your timezone.
- Two improvements were made to digest editor:
- Now it shows a latest discovered version near each packages and live update process started after a new package was added to the list.
- For packages with namespace
pythonwe are trying to figure out it's source repository and show autocomplete if something was found. We are taking this data from PyPi. Probably for other landguages/ecosystems we could do this as well.
- Fixed caching issue at digest page, when you've added a package, but digest seems not updated because of the cache.
- Fixed link to the package in digest email.
- Added cool feature — now our robot could download
and parse any Markdown or ReST changelog.
To force this behaviour, just add
http+before the source URL.
- Now authentication via Twitter or GitHub was added.
- Now users are able to add separate packages to the list and receive email with digest of recent changes.
This app provides an easy way to collect changelogs of different libraries. All information is gathered and presented in one format. There are at least two possibilities to collect changelog data.
Firstly, allmychanges.com tries to find a plaintext file with handwriten changelog. If it was found, we parse it and put to cache. If there is no handwritten changelog, our robot applies all his intelligence, to extract library's versions from VCS history and tie each commit message to particular version. This is not very accurate approach, but it is better than nothing.
First approach works with any kind of repository be it python, ruby or even java, whereas as second approach needs more insights what is code about. Right now, second approach only works with python modules, whose setup.py uses setuptools.
There are plenty directions for approvement. Stay tuned.